LMI Cuts Federal Tech Security Approvals to Six Weeks via Iron Sled
Ai4
Commercial vendors can bypass multi-year federal procurement delays by self-funding working prototypes before government requests appear and enforcing automated security checks on every code commit.
## Background
Published on YouTube by Ai4 on August 28, 2025, this session features LMI leaders Jared Summers and Alex Adamczyk discussing how to bridge the gap between commercial development velocity and federal mission requirements. According to the video description, the presentation outlines actionable strategies for designing scalable, compliant commercial-grade AI and emerging tech for government deployment without compromising mission integrity.
## TL;DR
Jared Summers and Alex Adamczyk break down why deploying commercial software into the federal government is hindered by centralized risk governance and complex compliance frameworks rather than a lack of private-sector innovation. To overcome these hurdles, they introduce LMI's three-pillar framework: innovating ahead of government procurement cycles through hands-on internal testing and live software demonstrations, building ATO-ready code on day one using automated DevSecOps platforms like Iron Sled, and designing modular, reusable architectures such as the NatEx sensor mesh. Their approach demonstrates how proactive engineering and rigorous security integration can slash deployment and certification timelines from years down to weeks.
## Atmosphere & style
The presentation carries a pragmatic, authoritative, and battle-tested tone, enriched by real-world defense leadership anecdotes and hard-earned institutional scars. Jared and Alex share an easy, collaborative dynamic, speaking directly to technical and mission leaders with clear-eyed realism about government acquisition friction. It feels candid, engaging, and energizing—moving away from theoretical defense procurement jargon toward actionable engineering rigor, capped off with lighthearted banter and crowd interaction.
## Key takeaways
- [00:04](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=4s) The primary divergence between commercial innovation and federal procurement is risk governance: private enterprises have the autonomy to accept risks, while federal agencies face centralized, non-negotiable compliance checklists.
- [01:22](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=82s) Federal software delivery requires passing an extensive gauntlet of compliance standards—including CMMC, NIST, FedRAMP, and Authority to Operate (ATO) approvals—before deploying into secure enclaves.
- [01:22](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=82s) Operational AI in defense settings involves lethal workflows and kinetic effects, creating a vastly higher risk profile than standard enterprise AI applications.
- [01:22](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=82s) LMI addresses the federal deployment gap using a three-pillar framework: innovate ahead of need, build ATO-ready architectures, and design for modular repeatability.
- [03:17](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=197s) Federal buyers are increasingly rejecting protracted multi-year development cycles that risk obsolescence, demanding proven, functional capabilities upfront.
- [03:17](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=197s) Dogfooding emerging technologies internally allows vendors to demonstrate live, interactive software workflows and instant policy impact analysis instead of theoretical white papers.
- [06:51](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=411s) Developers must account for cloud enclave parity and classification restrictions from day one, as many commercial cloud services do not exist in specialized federal environments.
- [06:51](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=411s) Enforcing automated DevSecOps gates and Software Bills of Materials (SBOMs) on every code commit can drastically compress ATO timelines from years to weeks.
- [09:57](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=597s) Cross-agency challenges—such as secure edge IoT data transmission—can be solved efficiently by building hardened, reusable backbones like NatEx with plug-and-play modular endpoints.
- [12:43](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=763s) Commercial technology can scale rapidly to federal mission speeds when compliance, supply chain security, and modularity are engineered directly into the codebase from inception.
## Detailed walkthrough
### [00:04](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=4s) Introduction: The Commercial vs. Federal Risk Dilemma
Jared Summers, Chief Technology Officer at LMI, opens the session alongside Alex Adamczyk, Vice President of Advanced Analytics & AI and lead of LMI's rapid prototyping team. LMI operates as a next-generation technology solutions provider dedicated to bringing commercial-grade, mission-ready capabilities to federal agencies on day one. Setting the stage for the presentation, they address a fundamental systemic question: why is it so notoriously difficult to translate commercial software development speed into federal procurement and deployment?
The core divergence between the commercial and public sectors does not stem from a lack of regulations or risks in private enterprise. Rather, it comes down to risk ownership and autonomy. Commercial organizations have the latitude to decide which risks they are willing to accept, balance, or bypass based on their business strategy. In stark contrast, federal risk governance is heavily centralized and bound to strict, non-negotiable compliance checklists. This structural lack of flexibility dramatically slows down the adoption of emerging technologies compared to the private sector.
[01:22](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=82s) The Federal Alphabet Soup and Kinetic AI Realities
Navigating federal software delivery requires confronting an intimidating "alphabet soup" of compliance frameworks: CMMC, NIST, FedRAMP, ATOs (Authority to Operate), and closed or air-gapped networks. Each framework represents an exhaustive gauntlet of third-party audits, strict technical certifications, and formal approvals that must be satisfied before any capability can touch a federal enclave.
Jared highlights the stakes by drawing on his previous role as CTO of the U.S. Army's 18th Airborne Corps—a three-star operational command encompassing roughly 100,000 soldiers that serves as the nation's immediate response force, capable of deploying battalion-sized elements anywhere in the world within hours. While deploying artificial intelligence into operational workflows like fire support, target intelligence, and critical combat tasks yielded massive efficiency gains, it also introduced life-or-death risk. Integrating AI into lethal workflows that produce "kinetic effects" carries a vastly different risk profile than deploying a standard commercial enterprise chatbot for routine human resources inquiries.
```mermaid
flowchart TD
accTitle: LMI Three Pillar Federal Deployment Framework
subgraph P1["Pillar 1: Innovate Ahead of Need"]
A1["Identify Industry & Academic Trends"] --> A2["Internal Dogfooding & Stress Testing"]
A2 --> A3["Live Demonstrations Over White Papers"]
end
subgraph P2["Pillar 2: Build ATO-Ready"]
B1["Understand Target Enclave Constraints"] --> B2["Iron Sled DevSecOps Platform"]
B2 --> B3["Pre-Approved Containers & Continuous Scans"]
end
subgraph P3["Pillar 3: Design for Modularity"]
C1["Identify Cross-Agency Problem Patterns"] --> C2["Build Core Secure Backbone (NatEx)"]
C2 --> C3["Plug-and-Play Edge Modules"]
end
P1 --> P2 --> P3
```
To bridge this divide without compromising security, LMI developed a three-part framework:
- **Innovate ahead of need**: Build and validate capabilities before formal government acquisition cycles catch up.
- **Build ATO-ready**: Integrate security, compliance, and supply chain governance directly into the codebase on day one.
- **Design for modularity and repeatability**: Construct reusable core architectures that scale across disparate agencies rather than building bespoke, one-off solutions.
[03:17](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=197s) Pillar 1: Innovate Ahead of Need
Federal customer expectations are shifting rapidly. Jared draws an analogy to consumer technology: no smartphone user would pay for an application upfront and wait two to five years for delivery, knowing full well that requirements and underlying technologies would be completely obsolete by launch. Federal agencies are adopting the same mindset; they no longer want to fund protracted multi-year development cycles in the vague hope of eventually receiving usable software.
Historically, federal contracting allowed vendors to submit speculative white papers outlining an approach and win $100 million awards without demonstrating working software. Alex explains that LMI flipped this dynamic on its head, particularly with generative AI. Long before federal agencies issued RFPs or allocated budget line items, LMI tracked academic and commercial breakthroughs and spent over a year building generative AI platforms internally. They tested and refined these tools across thousands of their own employees, gathering real-world feedback to ensure production resilience.
When LMI engaged prospective federal clients, they abandoned PowerPoint presentations and theoretical white papers entirely:
- **Live System Workflows**: Teams ingested real, publicly available agency documents directly into live software during the meeting.
- **Acquisition Automation**: Demonstrated end-to-end procurement and contracting automation workflows tailored to an agency's active purchasing needs in less than an hour.
- **Instant Policy Impact Analysis**: Processed executive orders released that very morning to immediately show clients how their internal policies would be impacted.
- **Art of the Possible**: Allowed clients to interactively query and challenge the platform in real time, turning speculative six-month pilot discussions into immediate day-one mission engagements.
[06:51](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=411s) Pillar 2: Build ATO-Ready with Iron Sled
Building for government deployment means knowing "the rules of the game before you start coding." Developers must account for the specific restrictions of target environments—such as classification levels and cloud enclave parity. Commercial cloud services native to commercial AWS, Microsoft Azure, or Google Cloud are frequently absent from specialized or classified federal enclaves; building dependencies on missing services guarantees failure on day one. Furthermore, modern federal mandates require strict Software Bills of Materials (SBOMs) to track all open-source components and mitigate software supply chain vulnerabilities.
To solve the historical bottleneck where prototyping took months but achieving an ATO took an additional 12 to 24 months, LMI created **Iron Sled**, their internal DevSecOps platform. Iron Sled enforces continuous compliance:
- **Pre-Approved Base Containers**: Every developer starts from a vetted, pre-approved container environment.
- **Automated Commit-Level Checks**: Recreates every down-stream ATO security gate on every single code commit and push.
- **Engineering Accountability**: Forces software engineers to evaluate package provenance, vulnerability profiles, and architectural impact before merging code.
This upfront rigor yielded dramatic operational results: when deploying a new capability to a federal customer, LMI presented a comprehensive SBOM, a clean vulnerability report, and a pre-packaged security dossier. The agency's authorizing officials had only three questions, slashing the ATO approval timeline from years down to just six weeks.
[09:57](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=597s) Pillar 3: Modularity, Repeatability, and the NatEx Sensor Mesh
The final pillar requires avoiding bespoke, single-use solutions in favor of modular, reusable platforms that address shared government problems. By engaging across defense and civilian spaces, industry partners can identify common structural requirements and engineer adaptable backbones.
Alex and Jared point to LMI's **NatEx** platform as a prime case study:
- **Core Commonality**: Across the Navy, Army, and federal healthcare sectors, customers consistently struggled with securely transmitting Internet of Things (IoT) edge data back to centralized command networks for decision-making.
- **Hardened Backbone**: LMI took mature, commercially available technology used for decades and hardened it to meet rigorous federal security and ATO standards.
- **Plug-and-Play Edge Adaptation**: Rather than re-engineering the entire system for every customer, NatEx serves as a secure central backbone. End-point hardware—whether RFID scanners, environmental sensors, or specialized tactical devices—simply plugs in as modular components.
- **Rapid Scaling**: Subsequent customer deployments require only minor configuration tweaks at the edge module layer, drastically reducing cost and deployment timelines across defense and civilian applications.
[12:43](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=763s) Conclusion and Audience Q&A
Jared and Alex conclude the session by reiterating that commercial innovation can successfully operate at mission speed within the federal government when vendors invest ahead of need, build ATO-ready architectures upfront, and design for modular repeatability.
Acknowledging the "lot of scars" earned through years of navigating federal deployment barriers, they invite attendees to continue the technical discussion at LMI's booth (Booth 525) or via lmisolutions.com, wrapping up the presentation with lighthearted banter while tossing company T-shirts into the audience.
## Notable quotes
> [00:04](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=4s) why is it so notoriously difficult to translate commercial software development speed into federal procurement and deployment? — Jared Summers
> [01:22](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=82s) alphabet soup — Jared Summers
> [01:22](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=82s) kinetic effects — Jared Summers
> [01:22](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=82s) Innovate ahead of need — Jared Summers
> [01:22](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=82s) Build ATO-ready — Jared Summers
> [01:22](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=82s) Design for modularity and repeatability — Jared Summers
> [03:17](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=197s) Art of the Possible — Alex Adamczyk
> [06:51](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=411s) the rules of the game before you start coding — Jared Summers
> [12:43](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=763s) lot of scars — Jared Summers
## Mentioned
Alex Adamczyk, Jared Summers, LMI, Ai4, 18th Airborne Corps, CMMC, NIST, FedRAMP, Authority to Operate (ATO), Iron Sled, NatEx, Amazon Web Services (AWS), Microsoft Azure, Google Cloud, U.S. Army, U.S. Navy
## Video description
Alex Adamczyk, VP, Advanced Analytics & AI, LMI
Jared Summers, CTO, LMI
Abstracts
Commercial innovation moves fast—fueled by agile development, cloud-native platforms, and rapid iteration. But federal missions demand security, scale, and reliability. In this session, Jared Summers and Alex Adamczyk will share how tech companies can design for scalability and extensibility from day one to meet the unique challenges of government deployment. They’ll explore real-world strategies for adapting commercial-grade AI and emerging technologies to federal use, from navigating compliance hurdles to accelerating adoption without compromising mission integrity. Join us for a candid conversation on delivering innovation at mission speed.
LMI通过Iron Sled将联邦技术安全审批缩短至六周
商业供应商可通过在政府提出需求前自筹资金开发可用原型，并在每次代码提交时执行自动化安全检查，从而避开长达数年的联邦采购延误。
Jared Summers与Alex Adamczyk深入剖析了阻碍商业软件部署至联邦政府的核心原因在于集中的风险治理与复杂的合规框架，而非私营部门缺乏创新。为克服这些障碍，他们介绍了LMI的三支柱框架：通过切实的内部测试和现场软件演示赶在政府采购周期之前开展创新；利用Iron Sled等自动化DevSecOps平台从第一天起就构建符合ATO标准的代码；以及设计诸如NatEx传感器网格等模块化、可复用的架构。他们的方案表明，前瞻性的工程设计与严格的安全集成能够将部署和认证周期从数年大幅缩短至数周。
商业创新与联邦采购的根本分歧在于风险治理自主权：商业企业可自主权衡风险，而联邦机构必须面对 CMMC、NIST、FedRAMP 等集中且严格的合规清单，且国防场景中涉及杀伤效应的 AI 风险极高。 (0:04)
联邦客户正逐步摒弃耗时数年且易面临技术过时的采购模式；LMI 通过在内部数千名员工中率先测试生成式 AI 工具，以实时文档处理与采购自动化现场演示替代了传统的概念白皮书。 (3:17)
为解决原型开发仅需数月但获批 ATO（运行授权）耗时两年的瓶颈，LMI 推出 DevSecOps 平台 Iron Sled，在每次代码提交时强制执行安全门禁与 SBOM 审查，将 ATO 审批周期大幅压缩至六周。 (6:51)
基于 NatEx 边缘数据平台，LMI 展示了模块化架构的跨机构复用价值：通过统一加固的中央骨干网络与即插即用的边缘传感器模块，低成本解决了陆军、海军及联邦医疗系统的通用数据回传难题。 (9:57)
Reading summary
Deploying commercial software into federal agencies is hindered by rigid, centralized risk governance rather than technical feasibility alone, according to LMI's Jared Summers and Alex Adamczyk. To bridge this divide, they argue that contractors must anticipate agency needs proactively, integrate compliance gates directly into development, and design modular systems for cross-agency reuse.
Divergent Risk Models and Kinetic AI Realities (0:04)
The primary barrier to adopting commercial technology in the federal sector is the structure of risk ownership. While commercial enterprises possess the latitude to accept or bypass risks based on strategic goals, federal agencies are governed by centralized, checklist-driven compliance frameworks, including CMMC, NIST, FedRAMP, and formal Authority to Operate (ATO) reviews on closed or air-gapped networks.
These constraints are magnified in defense contexts where failure carries life-or-death consequences. Drawing on his background as CTO of the U.S. Army's 18th Airborne Corps, Summers notes that deploying artificial intelligence to support fire coordination, targeting, and combat workflows presents a fundamentally higher risk profile than standard enterprise applications, necessitating strict vetting before software touches an operational enclave.
Anticipating Needs with Validated Prototypes (3:17)
Federal procurement expectations are shifting away from speculative, multi-year contracts that risk delivering obsolete systems. Historically, contractors could secure significant awards using theoretical white papers. In response to changing demand, LMI invested over a year developing and dogfooding internal generative AI platforms across thousands of employees before agencies issued formal solicitations.
Rather than presenting theoretical proposals, LMI used live demonstrations during customer meetings to showcase immediate capability. These sessions included ingesting public agency records, demonstrating end-to-end purchasing workflows in under an hour, and executing real-time policy impact analyses on executive orders released that morning to prove functional readiness on day one.
Compressing ATO Timelines via DevSecOps (6:51)
Achieving an ATO has historically stalled deployments for 12 to 24 months, particularly when software relies on commercial cloud services absent from classified or isolated federal environments. Modern compliance mandates also require detailed Software Bills of Materials (SBOMs) to secure software supply chains against component-level vulnerabilities.
To address this delay, LMI instituted Iron Sled, an internal DevSecOps platform that enforces continuous compliance from inception. By providing pre-approved container baselines and running automated security evaluations on every code commit, the platform forces engineers to address supply chain risks prior to merging code. In one highlighted agency deployment, supplying a complete SBOM and automated security dossier reduced the authorizing officials' review to three questions and cut the approval process to six weeks.
Modular Architecture Across Defense and Civilian Sectors (9:57)
To prevent costly, one-off engineering efforts, software vendors must identify shared requirements across disparate agencies. Summers and Adamczyk cite edge telemetry as a widespread challenge: the Navy, Army, and federal healthcare organizations all require secure methods to transmit Internet of Things (IoT) data back to central command systems.
Through its NatEx platform, LMI hardened mature commercial technology into a secure core architecture that meets federal compliance standards. By serving as an adaptable backbone, the system allows diverse hardware—such as RFID scanners, environmental sensors, or tactical devices—to attach as plug-and-play modules, enabling multi-agency deployment with minimal edge-level configuration changes.
Summers and Adamczyk maintain that the lag in federal technology adoption can be overcome if vendors abandon speculative proposals in favor of internally tested platforms, automated compliance pipelines, and modular designs suited for multi-agency application.
## Background
本内容由 Ai4 频道于 2025 年 8 月 28 日发布在 YouTube 平台上。视频记录了 LMI 首席技术官 Jared Summers 与高级分析与人工智能副总裁 Alex Adamczyk 的主题演讲，探讨科技企业如何从研发第一天起就针对可扩展性与可扩展性进行设计，以克服联邦政府在安全、合规与部署上的独特挑战，将商业级 AI 与新兴创新技术快速带入政府任务中。
## TL;DR
演讲围绕如何打破商业创新与联邦政府采办部署之间的壁垒展开。两位主讲人针对传统政府采购耗时长、合规流程繁复的痛点，提出了三大破局策略：超前于客户需求进行自筹研发与内部实测（Innovate ahead of need）、构建从第一天起就符合 ATO 授权标准的 DevSecOps 流程（Build ATO-ready）、以及拒绝孤岛式定制并转向高复用性的模块化设计（Design for modularity）。通过分享生成式 AI 现场演示、IronSled 合规平台实现 6 周闪电获批 ATO、以及为海军打造的 NEX 物联网骨干网等实战案例，系统阐述了如何将军规级安全硬化与敏捷商业模式相结合，实现“交付第一天”即投入实战。
## Atmosphere & style
整场演讲风格干练、务实且充满实战底气，两位讲者配合默契，将严肃的国防合规与技术架构讲得生动透彻。Jared 结合其在第 18 空降军的服役经历，将合规审查的“字母汤”与生死攸关的战术决策相连接，带来了强烈的代入感与现场说服力；Alex 则以坦诚幽默的语调自嘲过往踩坑留下的“伤疤”，让分享兼具洞察力与亲和力。现场氛围在严谨硬核的技术剖析与轻松互动的抛洒周边 T 恤中自然切换，展现出工业界老兵在突破官僚壁垒后的自信与洒脱。
## Key takeaways
- [00:04](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=4s) 私营企业拥有自主评估与承担风险的弹性，而联邦政府实行高度中心化、对照合规清单的审查机制，极大延缓了新技术落地节奏。
- [00:04](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=4s) CMMC、NIST、FedRAMP、ATO 以及物理隔离密闭网络等严苛的合规壁垒（“字母汤”），是横亘在商业创新与联邦任务之间的主要障碍。
- [00:04](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=4s) 在紧急反应部队与动能杀伤等高危作战工作流中引入 AI，其安全合规要求与承担的风险同日常行政用例有着本质区别，绝无试错容限。
- [02:49](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=169s) 破局传统采办拖延的三大支柱为：超前需求进行创新（Innovate ahead of need）、构建即符合 ATO 标准（Build ATO-ready）、按模块化与可复用性设计（Design for modularity and repeatability）。
- [02:49](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=169s) 联邦客户对敏捷速度的渴望空前强烈，不再愿意为长达 2 至 5 年的研发周期提前买单，避免交付时系统已落后过时。
- [03:58](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=238s) 颠覆传统依赖概念白皮书拿大合同的模式，LMI 提前逾一年自筹资金研发 GenAI 平台，并通过内部数千名员工深度试用打磨出商用级产品。
- [03:58](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=238s) 超前研发使对接动态从被动投标转为实机展示，能够在 1 小时会议中现场导入真实政务文件演练自动化采购与行政令比对，大幅加速采纳决策。
- [06:51](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=411s) 商业公有云原生 API 绝大多数未部署在政府涉密飞地，过度依赖公有云特有服务会导致系统迁移至涉密环境时瞬间瘫痪。
- [06:51](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=411s) 针对开源组件与模型的软件物料清单（SBOM）已成硬性审查指标，“原型做完再补合规”的旧方法往往导致 12 至 24 个月的严重滞后。
- [06:51](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=411s) LMI 打造 DevSecOps 平台 IronSled，基于官方预核准基础容器开发，每次代码 Push 均自动触发全套 ATO 审计，实现原生合规。
- [09:09](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=549s) 将安全审查前置融入日常工程文化，主动交付详尽 SBOM 与漏洞风险自评估，可将传统漫长的 ATO 获批流程奇迹般缩短至 6 周。
- [09:56](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=596s) 坚决摒弃为单个客户定制孤岛式系统的作坊模式，应横向提炼跨军种与跨部门的底层通用诉求，沉淀标准化核心能力。
- [09:56](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=596s) 以 NEX 物联网平台为例，将核心网络与安全通道作为骨干框架，各类异构传感器仅作为插件接入，使后续客户部署边际成本大幅降低。
- [11:42](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=702s) 采用在商业市场上历经数十年验证的成熟技术做底层，再进行军规级加固与安全硬化，能有效兼顾技术成熟度与联邦防御标准。
- [11:42](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=702s) 面向政府的任务交付完全可以实现“商业化运作”，依托超前投资、安全前置与模块化复用，能够在交付的第一天就将可用能力推向一线。
## Detailed walkthrough
### [00:04](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=4s) 联邦部署困境与安全合规的“字母汤”
在演讲开场，LMI 首席技术官 Jared Summers 与负责快速原型设计团队的 Alex Adamczyk 共同介绍了 LMI 作为联邦政府下一代技术解决方案提供商的定位，其核心使命是在“第一天”（Day 1）就为客户交付具备商用就绪（commercial-ready）能力的产品，帮助客户实现兼顾速度与精度的业务推进。他们提出了一个直击痛点的核心问题：为什么将商业模式与商业技术引入政府采办与部署体系会如此艰难？
Alex 指出，商业私营部门并非没有监管、问题或风险，但私营企业拥有高度的自主权去决定“承担哪些风险”以及“规避哪些风险”。相比之下，联邦政府内部的风险管理高度中心化，任何行动都必须逐项对照极为严格的合规清单。这种缺乏自主弹性的官僚审查机制极大拖慢了技术落地的节奏。
Jared 进一步列举了横亘在技术与部署之间的合规壁垒——CMMC、NIST、FedRAMP、ATO（授权运行许可）以及物理隔离的密闭网络（closed networks）。对于在联邦领域工作的人而言，这一连串令人眼花缭乱的“字母汤”（alphabet soup）代表着一长串必须通过的第三方认证、安全核对表与审批流程。
Jared 分享了自己此前担任美国陆军第 18 空降军（18th Airborne Corps）CTO 的切身经历。作为一支拥有约 10 万名士兵的三星级作战司令部，第 18 空降军是国家的“紧急反应部队”（Immediate Response Force），其任务要求在数小时内将营级作战力量投送至全球任何角落。在任期间，Jared 推动将 AI 深度嵌入火力支援（fire support）、情报研判等任务关键型工作流中：
> “如果我将 AI 加入到具有动能杀伤效果（kinetic effects）的作战工作流中，这与把 AI 仅用作人力资源（HR）流程的聊天机器人相比，其性质与风险承担完全不可同日而语。这就是合规与安全真正重要的地方。”
- **私营企业 vs 联邦政府风险机制**：私营部门具备灵活的风险评估与取舍权限，联邦政府则实行中心化管控与死板的合规清单机制。
- **美军第 18 空降军规模与任务**：约 100,000 名官兵，承担全球数小时内紧急空投部署的战略战备任务。
- **核心合规壁垒**：涵盖 CMMC、NIST 框架、FedRAMP 认证、ATO 授权及隔离密闭网络要求。
```mermaid
flowchart TD
accTitle: 商业与联邦环境技术部署路径对比
subgraph Commercial ["商业私营部门部署模式"]
C1["敏捷开发与快速迭代"] --> C2["企业自主风险权衡"]
C2 --> C3["快速上线投入市场"]
end
subgraph Federal ["联邦任务环境合规壁垒"]
F1["技术原型开发"] --> F2["合规字母汤审查<br/>(CMMC / NIST / FedRAMP)"]
F2 --> F3["ATO 授权与密闭网络审批"]
F3 --> F4["耗时数年完成部署上线"]
end
```
---
### [02:49](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=169s) 破局之道：重塑联邦采购的三大核心支柱
面对传统采购长达数年的拖延，Jared 与 Alex 提出了经过实战验证的破局三步法：**超前需求进行创新**（Innovate ahead of need）、**构建即符合 ATO 标准**（Build ATO-ready）、以及**按模块化与可复用性设计**（Design for modularity and repeatability）。
随着近年来几届政府对现代化转型的持续提速，联邦客户对速度的渴望空前强烈。Jared 用现代智能手机应用做了一个通俗的类比：没有任何消费者会为一款尚未开发完成的手机 App 提前付款，然后苦等 2 到 5 年才拿到下载权限——因为在这漫长的等待期内，用户的需求与外部环境早已发生了翻天覆地的变化。如今的联邦政府同样不再愿意预先支付漫长的研发成本，苦等一个最终交付时可能已经过时落伍的系统。
---
### [03:58](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=238s) 策略一：超前需求进行创新（Innovate Ahead of Need）
Alex 回顾了政府传统的采办逻辑：过去承包商仅凭一份构想白皮书，描绘一下解决思路，就能轻松拿下高达 1 亿美元的巨额合同，然后花费数年时间慢慢开发系统，甚至在签约时连任何实际技术原型都不需要演示。但在生成式 AI（GenAI）技术日新月异的今天，LMI 彻底颠覆了这种传统模式。
当生成式 AI 刚在学术界与工业界崭露头角时，没有任何政府客户提出相关需求或发布采办预算（由于联邦预算周期的滞后性，客户无法即时调整资金）。LMI 没有坐等客户立项，而是自筹资源提前投入了一年多的时间，在内部搭建起符合联邦安全标准的生成式 AI 平台。LMI 率先将系统部署给自己内部的数千名员工使用，经过全员的大规模“吃狗粮”（dogfooding）测试与高强度反馈迭代，在产品真正推向客户之前，就已经打磨成了成熟可靠的商用级软件。
这种超前研发彻底改变了销售与对接的现场交流动态。当 LMI 的团队与政府机构坐下来开会时，他们没有展示任何空洞的概念 PPT 或白皮书，而是直接打开运行中的软件平台，现场使用该机构的公开文件展开实时演练：
- **实时采办自动化**：现场抓取客户正在采购的项目文件，在短短一小时内共同构建出一套采购自动化工作流。
- **人事与组织流程**：现场演示 HR 业务场景的智能化梳理。
- **最新政策应对分析**：直接导入当天早晨刚刚颁布的总统行政令（Executive Order），系统瞬间对比该机构的现行政策，现场展示政策调整建议。
Jared 强调，这种“把成熟产品直接带入市场驱动采纳”的逻辑与商业世界毫无二致。只要证明产品是真实可用、并且第一天就能直接采购的，联邦政府就会迅速买单。
- **传统采办弊端**：仅凭一份白皮书拿下 1 亿美元合同，研发周期长达数年，交付即落后。
- **LMI 内部验证**：超前需求研发逾 1 年，在数千名内部员工环境中完成实际运行与反馈闭环。
- **现场实机演示**：在 1 小时的会议中，直接接入真实数据完成行政令比对与采购自动化工作流验证。
---
### [06:51](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=411s) 策略二：构建即符合 ATO 标准（Build ATO-Ready 与 IronSled 平台）
Jared 与 Alex 指出，在敲下第一行代码之前，开发团队必须彻底吃透目标部署环境的游戏规则与边界限制。一个常见的重大误区是：主流云厂商（如 AWS、Google Cloud、Microsoft Azure）的商用公有云原生服务，绝大部分并未全部部署在政府专有飞地（Government Enclaves）或涉密云环境中。如果技术产品强依赖于某些特定的公有云原生 API，一旦迁移至政府环境就会瞬间瘫痪。
此外，供应链安全正面临前所未有的严审，联邦政府已强制推行软件物料清单（SBOM, Software Bill of Materials）。系统使用了哪些开源模型与第三方依赖库？这些组件源自何处？由谁提供？供应链的任何一处潜在风险都可能直接导致系统被一票否决。
Alex 坦言，LMI 过去的原型开发模式与业内其他人一样：先做原型、向客户演示、引发客户兴奋，然后再回过头来想办法塞进客户的生产环境。这种“事后补合规”的旧模式往往会导致 12 到 24 个月的漫长延误。
为了根治这一痛点，LMI 打造了名为 **IronSled** 的内部 DevSecOps 平台。IronSled 从最底层重构了研发流程：
1. **基线合规镜像**：开发者从第一天起，就必须使用经过官方安全预先核准的干净基础容器镜像（Clean Approved Containers）。
2. **提交即扫描**：代码库的每一次 Push 提交，都会自动触发并跑通后续 ATO 审查所需面对的全部严苛安全检查流程。
3. **强制防错机制**：从源头杜绝了任何不合规代码的滋生空间，使产品在原型阶段就原生具备 ATO 交付标准。
```mermaid
flowchart LR
accTitle: LMI IronSled 持续合规 DevSecOps 流程
subgraph IronSled ["IronSled DevSecOps 核心机制"]
A["预核准合规容器<br/>(Clean Approved Container)"] --> B["代码编写与组件引入<br/>(包含 SBOM 追踪)"]
B --> C["代码提交 (Push)"]
C --> D["自动化 ATO 全流程核查<br/>(安全扫描 / 策略审计)"]
D --> E["交付即符合 ATO 标准<br/>(ATO-Ready on Day 1)"]
end
```
- **云飞地（Enclave）服务缺失风险**：商业公有云（AWS/GCP/Azure）功能不能默认等同于政府涉密飞地可用功能。
- **软件物料清单（SBOM）**：严格追溯所有开源模型与外部代码组件的来源与供应链风险。
- **传统延误周期**：原型完成后再进行安全与合规改造，通常需耗费 12、18 乃至 24 个月。
- **IronSled 平台特性**：基于预核准容器构建，代码每次推送均执行自动化 ATO 标准合规审计。
### [09:09](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=549s) 安全前置的工程文化与 6 周闪电获批 ATO
工程团队必须在敲下每一次提交时，将安全思维深深根植于日常操作中。每当工程师在代码库点击“提交”（commit）时，脑海中都必须自发审视一系列关键问题：这个开源软件包来自哪里？项目到底该不该引入它？它是否存在已知漏洞？这种思维方式从根本上改变了软件架构的设计路径。如果团队愿意在前期做足功课，自始至终秉持“安全内置”的理念进行架构设计，后续的落地合规流程就会变得无比干净利落。
这种前期投入在面对联邦政府客户时带来了颠覆性的成效。以 LMI 将旗下另一款工具引入联邦政府的亲身经历为例，团队从与客户的初次接触交谈，到最终正式斩获 ATO（授权运营，Authority to Operate），前后仅仅耗费了整整 6 周时间。在联邦采购体系中，这一速度与过去的拖沓流程相比完全是天壤之别。之所以能如此高效，关键在于团队在前期就已经将所有的安全隐患清理完毕。当正式交付评审时，团队直接甩出一整套完整的物料清单与安全证明：“这是我们的 SBOM（软件物料清单），这是我们的风险评估报告，这是关于所有已知漏洞的健康证明，我们已经做好准备与你们一起走完流程。”最终，审查人员仅仅提出了大约三个问题便顺利过关，交付直接宣告完成。
- **极速 ATO 获批**：得益于前期的安全架构设计，从与客户首次沟通到拿下 ATO 仅耗时 6 周。
- **合规交付三件套**：主动向审查方提供详尽的 SBOM（软件物料清单）、风险自评估报告以及全量漏洞排查证明，将漫长的安全拷问精简至区区三个常规问题。
### [09:56](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=596s) 拒绝定制化：面向模块化与可复用性打造物联网骨干架构
在联邦业务中，企业必须像在商业领域那样，从第一天起就坚持面向“模块化”（modularity）与“可复用性”（repeatability）进行设计，坚决杜绝为单个客户定制孤岛式的专用解决方案（bespoke solutions）。作为业界承包商，最大的核心优势在于能够横向接触政府内部各个不同层级与部门，从而洞察他们共同面临的痛点。如果能够将这些纷繁复杂的诉求抽丝剥茧，提炼出跨部门真正通用的核心能力，并围绕该核心构建易于拼接的扩展模块，那么政府各机构采购这些能力的成本将大幅降低，部署速度显著提升，厂商自身也能取得更大的商业成功。
以 LMI 为美国海军打造的 NEX（或称 Natex）平台为例，该平台是一个专为物联网设备设计的传感器网状网络（sensor mesh）。在与客户沟通时，团队反复听到来自各方的共性诉求：他们急需一个安全可靠的物联网 Mesh 骨干框架。虽然在具体应用中，医疗部门、海军、陆军所挂载的物联网硬件形态天差地别，但它们底层都需要一条通用的中心骨干线路，以便安全地将数据从边缘端回传至核心决策层。通过抓住这一高频痛点，团队采用了高度模块化的设计方式，无论是射频识别（RFID）标签还是其他新型物联网传感器，都只需作为即插即用（plug-and-play）的外围插件接入到这一安全架构中。这种设计彻底告别了“面对每个新客户都拿出一张白纸从零开始”的作坊模式，使得第二位、第三位乃至第四位后续客户在部署时，只需在边缘微调适配单一模块，即可实现闪电交付。
```mermaid
flowchart LR
accTitle: NEX平台模块化物联网架构图
subgraph EdgeDevices["异构边缘设备 (即插即用)"]
D1["海军舰船传感器"]
D2["陆军战术物联网"]
D3["医疗 RFID 标签"]
end
subgraph CoreBackbone["NEX 安全通用骨干网"]
B1["即插即用接入适配层"]
B2["预置安全与合规通道"]
end
subgraph DecisionLayer["指挥决策核心"]
C1["边缘安全数据回传"]
C2["任务决策与数据分析"]
end
EdgeDevices --> B1
B1 --> B2
B2 --> C1
C1 --> C2
```
- **抽象核心公约数**：跨部门提取底层共性需求，避免陷入“针对单一客户需求重新发明轮子”的设计陷阱。
- **插件化快速适配**：基础骨干网络保持固化与高强度安全，仅在边缘针对 RFID 或专用硬件做极简调试，边际部署边际成本大幅下降。
### [11:42](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=702s) 超前投资与商业成熟技术硬化：三大成功法则总结
NEX 平台的诞生完美印证了本次分享所阐述的三项核心法则。首先是“先于需求进行超前投资”（invest ahead of need）：早在客户自己意识到真实需求之前，团队就已经洞察到市场机遇，提前启动平台研发，并前置推进各项安全认证流程；其次，团队选用了在商业市场上已经历经数十年检验的成熟技术，在此基础上进行军规级“硬化”（hardened）与安全强化；最后，通过“合规在先（ATO First）”与极高的模块化架构推向市场，使产品不仅局限于海军或国防部（DoD），而是能够快速响应整个客户群体的各种新兴用例。
在短短二十分钟的分享里，两位主讲人高度提炼了赋能 LMI 成功驾驭商业创新跨越到联邦任务的关键策略。他们强调，向政府交付解决方案并非只能慢工出细活，“你完全可以将政府交付流程商业化”。正是凭借超前投资、安全前置以及模块化重用这三板斧，团队才能帮助联邦客户以更高的精度和更快的速度迈进，在交付的第一天就将成熟、安全的技术交到一线最终用户手中。
演讲在轻松热烈的互动中走向尾声。主讲人感慨道，在过去几年摸索这套模式的历程中，团队走过了无数弯路，“身上留下了大把伤疤”（a lot of scars）。在向现场听众致谢并引导大家前往 525 号展台或登录官网深入交流之后，台上甚至现场抛投起周边 T 恤活跃气氛，引来全场阵阵欢笑与掌声。
- **三大核心策略矩阵**：
1. **超前投资**：在客户尚未形成明确采购需求前，预先完成技术验证与资质认证储备。
2. **商业技术硬化**：利用商业领域沉淀数十年的成熟底层技术，做针对性安全加固而非重起炉灶。
3. **模块化与合规在先**：以开箱即用的 ATO 资质和可拔插组件，实现全军种及民事机构的多场景泛化复用。
- **展台与联络指引**：会后交流设在场外 525 号展位，亦可通过 lmisolutions.com 了解进一步详情。
## Notable quotes
> [00:04](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=4s) 如果我将 AI 加入到具有动能杀伤效果（kinetic effects）的作战工作流中，这与把 AI 仅用作人力资源（HR）流程的聊天机器人相比，其性质与风险承担完全不可同日而语。这就是合规与安全真正重要的地方。 — Jared Summers
> [02:49](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=169s) Innovate ahead of need — Jared Summers & Alex Adamczyk
> [02:49](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=169s) Build ATO-ready — Jared Summers & Alex Adamczyk
> [02:49](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=169s) Design for modularity and repeatability — Jared Summers & Alex Adamczyk
> [09:09](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=549s) 这是我们的 SBOM（软件物料清单），这是我们的风险评估报告，这是关于所有已知漏洞的健康证明，我们已经做好准备与你们一起走完流程。 — Alex Adamczyk
> [11:42](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=702s) 你完全可以将政府交付流程商业化 — Jared Summers & Alex Adamczyk
> [11:42](https://www.youtube.com/watch?v=AvOGIWhXjEQ&t=702s) 身上留下了大把伤疤（a lot of scars） — Jared Summers
## Mentioned
Jared Summers, Alex Adamczyk, LMI, Ai4, 美国陆军第18空降军 (18th Airborne Corps), CMMC, NIST, FedRAMP, ATO (Authority to Operate), IronSled, SBOM (Software Bill of Materials), NEX (Natex), AWS, Google Cloud, Microsoft Azure, 美国海军 (US Navy), 美国国防部 (DoD)
## Video description
Alex Adamczyk, VP, Advanced Analytics & AI, LMI
Jared Summers, CTO, LMI
Abstracts
Commercial innovation moves fast—fueled by agile development, cloud-native platforms, and rapid iteration. But federal missions demand security, scale, and reliability. In this session, Jared Summers and Alex Adamczyk will share how tech companies can design for scalability and extensibility from day one to meet the unique challenges of government deployment. They’ll explore real-world strategies for adapting commercial-grade AI and emerging technologies to federal use, from navigating compliance hurdles to accelerating adoption without compromising mission integrity. Join us for a candid conversation on delivering innovation at mission speed.
[Source](https://www.youtube.com/watch?v=AvOGIWhXjEQ)
Source quotes
[0:04] Jared Summers: 如果我将 AI 加入到具有动能杀伤效果（kinetic effects）的作战工作流中，这与把 AI 仅用作人力资源（HR）流程的聊天机器人相比，其性质与风险承担完全不可同日而语。这就是合规与安全真正重要的地方。
[2:49] Jared Summers & Alex Adamczyk: Innovate ahead of need
[2:49] Jared Summers & Alex Adamczyk: Build ATO-ready
[2:49] Jared Summers & Alex Adamczyk: Design for modularity and repeatability
[9:09] Alex Adamczyk: 这是我们的 SBOM（软件物料清单），这是我们的风险评估报告，这是关于所有已知漏洞的健康证明，我们已经做好准备与你们一起走完流程。
[11:42] Jared Summers & Alex Adamczyk: 你完全可以将政府交付流程商业化
[11:42] Jared Summers: 身上留下了大把伤疤（a lot of scars）
Source quotes
[0:04] Jared Summers: why is it so notoriously difficult to translate commercial software development speed into federal procurement and deployment?
[1:22] Jared Summers: alphabet soup
[1:22] Jared Summers: kinetic effects
[1:22] Jared Summers: Innovate ahead of need
[1:22] Jared Summers: Build ATO-ready
[1:22] Jared Summers: Design for modularity and repeatability
[3:17] Alex Adamczyk: Art of the Possible
[6:51] Jared Summers: the rules of the game before you start coding
[12:43] Jared Summers: lot of scars
Original transcript
[0:04] Morning everyone. Jared Summers, CTO of
[0:07] LMI and I'm here with I am Alex Adamic.
[0:10] I lead our rapid prototyping team. And
[0:11] for those who uh haven't heard of LMI
[0:13] before, uh we are a next generation uh
[0:16] technology solutions provider for the
[0:17] federal government that bring
[0:19] commercialrade capabilities uh on day
[0:22] one that enable our customers to move
[0:24] faster with precision. So for the next
[0:27] 19 minutes, uh we're going to walk
[0:29] through a use case of how LMI has
[0:32] evolved over the last few years to
[0:34] address the changing needs of the
[0:36] government for commercial ready
[0:38] solutions on day one to help them move
[0:40] faster.
[0:42] And so with that, why is it so hard to
[0:44] apply a commercial approach to
[0:45] government procurement and deployment?
[0:49] And so it's not that the private sector
[0:52] doesn't have regulations, doesn't have
[0:53] problems, doesn't have risks, but in the
[0:55] private sector, you have way more
[0:57] control of the federal government about
[0:58] how you choose to make those risks.
[1:00] Which risk do you want to take, which
[1:01] ones do you not? There's so much more
[1:03] flexibility and choice on individual
[1:05] companies to make those decisions. In
[1:06] the federal government, so much more of
[1:08] that risk is centralized, and there's a
[1:10] checklist you need to go through to make
[1:11] sure you're able to do things. And so
[1:13] there's just a lot less flexibility in
[1:15] the federal government than there is in
[1:16] the commercial side around what risks
[1:18] can I take, what can I do, what can I
[1:19] not do that slows things down compared
[1:21] to the private sector.
[1:22] >> So CMMC, NIST, Fed ramp, ATOS, closed
[1:28] networks, uh for those who operate in
[1:30] the federal space, you understand what
[1:32] that alphabet soup means. Each one of
[1:34] those are a list of requirements and
[1:37] checklists and certifications and
[1:39] third-party uh approvals that you have
[1:41] to go to before you can ever deploy or
[1:43] bring a capability to your customer. And
[1:45] on the federal side, a great example in
[1:48] my previous role, I was the CTO of the
[1:49] 18th Airborne Corps. For those that
[1:51] don't speak Army, uh that is a a
[1:53] three-star operational command of about
[1:55] a 100,000 soldiers across the United
[1:58] States. And our mission and our role,
[2:00] the the reason the 18th Airborne Core
[2:02] Corps exists is we're the immediate
[2:04] response force for the nation. And so
[2:06] what that means is we can take a
[2:08] battaliansized element and project them
[2:11] anywhere around the world within hours.
[2:13] And when I was there, we were embedding
[2:16] and deploying AI into our operations to
[2:19] make us more efficient. So whether that
[2:21] was for fire support, intelligence, uh,
[2:24] and other mission critical tasks, we
[2:26] were leveraging AI to make us more
[2:28] effective and more efficient. But what
[2:30] that means is there's more risk. Because
[2:32] if I'm adding AI to my workflows that
[2:36] have kinetic effects, that's a lot
[2:38] different than if I'm adding AI to a
[2:42] chatbot for, you know, an HR process.
[2:45] And and that's why this matters.
[2:49] And so what we've learned over the last
[2:51] couple years is there is a better way.
[2:53] If you follow these three steps, you can
[2:55] get ahead of the problem, think through
[2:56] it so that when you bring your solution
[2:58] to the government, you can be a helpful
[3:00] partner and skip a lot of this long-term
[3:02] process.
[3:04] So the the three steps or three core
[3:06] approaches that we're going to talk
[3:07] about for the next few minutes are
[3:09] innovate ahead of need, build atto
[3:12] ready, and design for modularity and
[3:14] repeatability.
[3:17] So the government wants to move fast.
[3:19] We've actually seen this over the last
[3:21] few years, but in the recent
[3:22] administration, it has just accelerated.
[3:25] You know, take the analogy. None of us
[3:28] would pay for a product up front, an
[3:30] application on your phone that wasn't
[3:32] ready. Would you pay and then wait for
[3:34] two or 5 years for that capability to
[3:37] deliver your phone? Well, no. And also
[3:39] your needs and requirements are going to
[3:42] change so many times over the that that
[3:46] period that you're waiting. And the
[3:47] government is no different. They no
[3:49] longer want to wait. They no longer want
[3:51] to pay for that development in the hopes
[3:53] that one day they can get a capability
[3:55] that they can use and it not be
[3:56] outdated. And so that's what we're
[3:58] really talking about. So, as Jared said,
[4:00] for so many years, the way the
[4:02] government bought things was just
[4:03] completely different the way the
[4:04] commercial did. You could write a white
[4:06] paper, show up and say, "This is how I
[4:08] think I'm going to approach the
[4:09] problem." And you could $100 million
[4:11] contract off of a white paper with no
[4:13] demonstration, no technology, and then
[4:14] you would have years to go get it up and
[4:16] running. And LMI, we've flipped this on.
[4:18] We especially on generative AI where
[4:20] things are moving so fast. No one was
[4:22] asking for generative AI when we started
[4:24] building our technologies. We saw what
[4:26] was happening in industry. true saw was
[4:27] happening in academia and started
[4:29] building what does that need to look
[4:31] like inside the federal government from
[4:32] the security standpoint long before any
[4:35] customers were putting out RFPs or
[4:37] information like that because no one had
[4:38] any money planned ahead of time for this
[4:41] in their budgets and the budget cycle so
[4:42] much longer in the federal space that
[4:44] they can't just pivot on a dime and say
[4:46] I'm changing now I'm buying genai and so
[4:48] what that meant for us is we spent over
[4:50] a year building this technology for
[4:52] ourselves deploying it to our own our
[4:54] own employees testing it on ourselves So
[4:57] that when we brought it to that first
[4:58] customer, thousands of people had
[5:00] already used it, given us lots of
[5:02] feedback, changed how it worked. So we
[5:04] knew it was going to work for those
[5:05] customers. That made it so that the
[5:07] first time we showed up, we weren't
[5:09] telling them, "Here's what Gen AI is,
[5:11] and we think it's going to be useful for
[5:12] you, or here's a couple examples of what
[5:14] it look like." We could actually pull up
[5:15] the working software, sit down with
[5:17] them, and walk through their use cases
[5:20] live with them. Those conversations were
[5:22] completely different. We did not show up
[5:25] with a PowerPoint presentation or a
[5:26] white paper. We pulled here's the things
[5:29] you care about in your agency. Public
[5:30] documents live in our system. Ask
[5:34] questions. Let's build workflows. You
[5:35] want to show what an acquisition
[5:37] automation would look like. Here's
[5:39] things for something you're actually
[5:40] buying. Let's recreate that together
[5:42] with you in the span of an hour. You
[5:43] want to do HR processes? Let's do it.
[5:45] Let's You want to understand how your
[5:46] policies would change by the new
[5:48] executive orders? Great. Here's the
[5:49] executive order that came out this
[5:51] morning. Here's your policies. Let's
[5:53] just use all these tools in the meeting
[5:55] to see the art of the possible. And
[5:57] those conversations were completely
[5:59] different than, hey, 6 months from now,
[6:02] we'd love to have a working prototype
[6:03] with you for a pilot. And if you get
[6:05] ahead of that, have those conversations
[6:06] early with customers, you can have a
[6:08] much much more productive conversation.
[6:10] And you're ready on day one to actually
[6:11] solve a problem, not start getting
[6:13] started on it. And
[6:14] >> so when you think about it, it is
[6:16] actually no different than the
[6:17] commercial sector. You're building your
[6:19] products, you're getting ready to
[6:21] deploy, and you're bringing those to
[6:23] market to actually drive adoption and
[6:25] and acquisition. It's no different in
[6:27] the government. And so folks think it's
[6:30] different. It's not. The government has
[6:32] realized they're moving on with
[6:33] commercial solutions opening. You have
[6:35] to demonstrate the commercial ability.
[6:36] Your product has to be real and it has
[6:39] to be available for them to buy. And so
[6:41] if you invest in head of need just as
[6:43] you would in the commercial sector,
[6:44] you're going to see the same results in
[6:46] the federal sector. as long as you
[6:48] follow additional steps.
[6:51] So, so the first is there. The next is
[6:54] build atto ready. So, I already labeled
[6:57] off a whole host of CMC bed ramp atto's
[7:01] and everything else. But what does that
[7:03] really mean? That really means that
[7:05] understand the rules of the game before
[7:08] you start coding. Before you make the
[7:11] first line of code, understand what is
[7:13] allowable in your target environment.
[7:16] And that could be classification, that
[7:18] could be customer. Great example. All of
[7:21] the uh native services in whether it's
[7:23] AWS, GCP or or you know Azure, they're
[7:27] not all available in all of the
[7:29] government enclaves. So if you have
[7:31] built your product requiring native
[7:34] services that are not in the enclave
[7:36] that you want to deploy, your capability
[7:38] is not going to work on day one. It's
[7:41] not. And so understanding those
[7:43] restrictions and understanding what's
[7:45] available is critically important.
[7:47] Another is the government's not
[7:48] requiring sbombs. So that is actually a
[7:51] software build materials. What
[7:53] open-source models are you using and
[7:56] leveraging within your product and where
[7:58] do they come from and who is supplying
[8:00] them? Software software uh supply chain
[8:04] is an incredible real risk. uh and again
[8:07] it goes back to how is the solutions
[8:09] being provided and what is the level of
[8:11] risk that the federal government can
[8:12] take and so this has completely changed
[8:15] the way I do prototyping inside of
[8:17] organization. We used to do it the same
[8:19] way as everyone else did of build a
[8:20] prototype, demonstrate it works, show it
[8:22] to a customer, get them excited about
[8:23] it, and then start to figure out how to
[8:25] fit in their environment. And we kept
[8:26] running into years and years and years
[8:28] of well, how long is it going to take
[8:30] till I can I can use this? And if we
[8:31] haven't, so if you haven't done your
[8:32] security up front, that could be 12, 18,
[8:35] 24 months. And so instead, we built iron
[8:38] sled, which is our dev sec ops platform
[8:40] inside LMI that we built all of our
[8:41] tools on top of. And so what it does is
[8:44] it recreates all of the checks that
[8:47] you're going to have to go through later
[8:48] in the process for that at every time
[8:51] someone pushes code. So from the first
[8:53] day you create something, you start with
[8:55] a clean container that's already
[8:56] approved. And then every time someone
[8:58] wants to make a change, they have to run
[9:00] through that process. And so you can't
[9:02] build it the wrong way. You can't build
[9:04] something, show up, and find out you did
[9:06] it wrong. And what it does is it forces
[9:09] everyone on our engineering teams to
[9:11] think about security every time they hit
[9:13] submit. Every time they hit commit on
[9:15] any of that code, they're thinking
[9:16] through where did this package come
[9:18] from? Should I be using this? What
[9:20] vulnerabilities does it have? That
[9:22] changes how you design your
[9:23] applications. But if you do that work up
[9:26] front, you architect with security in
[9:28] mind, you're going to have a much much
[9:30] cleaner process. And so when we brought
[9:31] one of our other tools to the federal
[9:33] government because we had built in this
[9:34] secure way up front from our first
[9:36] conversation with a customer to
[9:38] receiving our ATO was 6 weeks which is
[9:41] completely different than the past the
[9:43] process we've had in the past because we
[9:45] already had all our security taken care
[9:46] of and we handed them here's our sbomb
[9:49] here's our risks here's our clean bill
[9:51] of health on all of our vulnerabilities
[9:53] we're ready to go through the process
[9:54] with you and they had like three
[9:55] questions and we were done.
[9:56] >> Awesome. And then so the next is design
[9:59] for modularity and repeatability just as
[10:01] you would in the commercial sector.
[10:03] Don't build bespoke solutions. Ensure
[10:06] that you're building for modularity and
[10:08] repeatability for the beginning. We have
[10:10] multiple great examples of that that we
[10:13] can reuse the building blocks to quickly
[10:16] change and respond to new use cases and
[10:18] expand to other customers. And one of
[10:21] your benefits from industry is you get
[10:23] to talk to lots of different parts of
[10:24] the government and find the
[10:25] commonalities in their problem. And if
[10:27] you can boil that back to what is the
[10:29] core thing many many of these people
[10:31] need and just build that and like get
[10:33] easy build modules around it then it's
[10:35] much faster and cheaper for all the
[10:37] governments to buy those and you're
[10:38] going to have a lot more success in
[10:39] this. And so we did this with our our
[10:41] NEX platform which is a sensor mesh used
[10:44] in the Navy where we kept hearing from
[10:46] all these different customers. They need
[10:48] a secure mesh framework for IoT devices.
[10:51] What those IoT devices looks completely
[10:54] different in healthcare, in the Navy, in
[10:56] the army, but they need some central
[10:59] backbone that they can hang these things
[11:00] off to get something to get data
[11:02] securely from the edge back to where
[11:04] it's need to make decisions. And so by
[11:06] boiling in, okay, we see this common
[11:09] problem over and over and over again,
[11:10] you build it such a modular way that
[11:12] whatever that RFID device is or whatever
[11:14] whatever thing you need to add that to
[11:16] the IoT, that's just a plug-and-play
[11:18] thing into the secure architecture. So
[11:20] the second customer, the third customer,
[11:21] the fourth customer has so much faster
[11:24] of a deployment process because they're
[11:25] just tweaking one little module at the
[11:27] end. Again, it's thinking a lot more
[11:29] about the architecture upfront to make
[11:31] it easy to redeploy this instead of
[11:34] approaching every single thing with a
[11:35] blank sheet of paper of this is how I'm
[11:37] going to solve this one customer's
[11:39] problem and the starting go over again
[11:40] when you get to that next customer.
[11:42] >> And uh Natex is a fantastic use case of
[11:44] everything that we just walked through
[11:46] the three the three items. The first is
[11:48] we saw the opportunity and we invested
[11:50] ahead of need. We started to build the
[11:53] platform. We started to work through the
[11:54] accreditation process before our
[11:56] customers even knew they need needed it.
[11:59] We took commercially available
[12:00] technology that has been used for
[12:02] decades uh and then hardened it, secured
[12:05] it and then brought it to market with
[12:07] the atto first and the modularity to
[12:09] rapidly respond and move to to various
[12:12] use cases across our entire customer
[12:14] base, not just the Navy or the DoD.
[12:17] And so you can commercialize government
[12:20] deployment. NAX was a great example. So,
[12:22] we only had 20 minutes to walk through
[12:25] three very simple key things that have
[12:28] enabled LMI to be incredibly successful
[12:30] to manage this this transition and keep
[12:33] pace with our customers to enable them
[12:35] to move faster and with precision to put
[12:38] the capability in the end user's hands
[12:40] on day one.
[12:43] And with that, thank you so much. We've
[12:46] gone through this process many, many,
[12:48] many times over the last couple years. A
[12:49] lot of scars. Please stop by outside.
[12:51] We'll be giving you questions, want to
[12:53] go dive deeper into any of this or stop
[12:55] by our booth 525 outside or find look us
[12:58] up at lmisolutions.com.
[12:59] Thank you so much.
[13:00] >> Anybody want a shirt?
[13:02] >> Yep. We're there. All right,
[13:03] >> boy.
[13:05] [Laughter]
[13:08] >> Oh man, I don't know if I can go that
[13:10] far.
[13:13] >> Almost.
[13:13] >> All right, please help me thank Jared
[13:15] and Alex.
[13:16] >> Thank you.
