Lessons from Project Warp Speed in How High-Velocity Partners Are Scaling AI in Manufacturing – with Emily Nguyen of Palantir Technologies

American manufacturers are pouring millions into keeping decades-old systems alive, even as the industry races toward AI-enabled operations. The result is a widening efficiency gap: firms with tightly integrated digital environments are accelerating output and cutting costs, while the majority remain stuck in fragmented ERP and production systems that drain resources and slow decision-making.

Maintaining and patching legacy systems in American manufacturing costs the average business $2.9 million, according to a SnapLogic survey of 750 IT decision-makers.

Despite the US leading global software by significant margins, manufacturers remain trapped in what Palantir’s Emily Nguyen calls “walled garden” systems — isolated ERP, MES (Manufacturing Execution Systems), and PLM (Product Lifestyle Management) platforms that consume financial resources.

This technological paradox has created a crisis of operational inefficiency: a study from the European Journal on Operational Research concluded that firms with automated, integrated systems achieve significantly greater cost and time efficiencies. Meanwhile, 70% of manufacturers still enter data manually, according to a report from the American Manufacturing Association.

The solution lies not in wholesale system replacement, but in deploying AI as a universal translator that bridges these technological silos. So claims Palantir from findings in its latest endeavor, Project Warp Speed, which features partnerships with advanced manufacturers in the process of inspecting their data governance infrastructures.

Emerj Editorial Director Matthew DeMello recently spoke with Emily Nguyen, Head of Industrials at Palantir Technologies, to discuss how Project Warp Speed approaches rapid and effective AI scale with various manufacturing operations. She breaks down the key values her team uses to drive successful AI adoption and shares relevant examples from real-life experience.

The following analysis of their conversation examines two key insights:

• Unifying fragmented manufacturing environments: Leveraging multi-modal AI and digital twins to seamlessly connect legacy systems that preserve institutional knowledge.
• Automating and optimizing factory operations: Deploying advanced computer vision and predictive analytics on the factory floor to accelerate quality control, preserve tribal knowledge, and proactively resolve bottlenecks.

Listen to the full episode below:

Guest: Emily Nguyen, Head of Industrials, Palantir Technologies

Expertise: Data Integration, Operational Intelligence, and Defense Analytics

Brief Recognition: Emily leads critical initiatives that drive enterprise-scale digital transformation across public and private sectors. At Palantir, she’s been instrumental in deploying data platforms that empower governments and Fortune 500 companies to make real-time, high-stakes decisions. Her work emphasizes infusing complex data infrastructure with accessible, actionable outcomes — especially in domains like national security, logistics, and healthcare.

Unifying Fragmented Manufacturing Environments

Nguyen’s most penetrating observation centers on what she terms the “walled garden” problem plaguing American manufacturing. These legacy systems — PLM, ERP, and MES platforms — weren’t designed with interoperability in mind.

The economic implications extend far beyond operational inefficiency. According to Nguyen, companies with rigid systems can’t pivot quickly enough to respond to market signals. They can’t reallocate resources dynamically, can’t adjust production schedules based on real-time customer needs, and can’t optimize inventory across multiple facilities.

The COVID-19 pandemic exposed these vulnerabilities with brutal clarity. Nguyen recounts the chaos in the meat supply chain, where companies swung wildly between extremes:

“At times we had way too much supply. So you saw companies depopulating their livestock. Then we had too little and there was discussion about using the Defense Production Act to get it back on track.”

– Emily Nguyen, Head of Industrials at Palantir Technologies

The situation wasn’t merely supply chain disruption — it was a systematic failure of inflexible automation systems.

Companies invest millions in sophisticated ERP systems, only to abandon them for Excel spreadsheets when flexibility becomes critical. The consequences ripple through every manufacturing operation. Indeed, there exists a large disconnect between Silicon Valley’s algorithmic capabilities and the realities of factory floor implementation.

Project Warp Speed, continues Emily, addresses this fragmentation through ontology-based architectures that create unified data models spanning design, engineering, manufacturing, supply chain, and sustainment operations.

Nguyen highlights that this technical framework leverages large language models (LLMs) as universal translation layers, enabling semantic interpretation of unstructured manufacturing data while maintaining compatibility with legacy system APIs. This architecture eliminates the traditional requirement for extensive data standardization projects.

At L3Harris manufacturing facilities, Nguyen says, this manifests as comprehensive program management that aggregates real-time data from disparate sources — inventory levels, program schedules, engineering drawings, financial metrics — into unified executive dashboards.

The underlying AI systems continuously correlate cross-functional data streams, identifying risk patterns and opportunity signals that would remain invisible within traditional siloed architectures.

Automating and Optimizing Factory Operations

Nguyen then segues into the kinds of principles that typically guide successful AI deployment:

• Relentless pursuit of outcomes: Put primacy on outcomes, focusing on the global optimum rather than point solutions
• First principles thinking: Be methodical about what you need, why, and when; reject “this is how it’s always done” thinking
• Speed and urgency: Prioritize fast iterations; learning accelerates innovation

In manufacturing industries especially, Nguyen emphasizes that AI applications must preserve organizational knowledge. She describes meeting workers at L3Harris who’ve worked for 30-40 years and hold large troves of “tribal knowledge” — insights that are not captured in instruction manuals, or knowledge that makes the difference between mediocre and exceptional performance.

Emily goes on to explain that “our hope is to provide an AI assistant that allows us to begin to capture tribal knowledge and make it more available to the broader workflow.”

Rather than losing decades of accumulated expertise when employees retire, companies can now encode that institutional knowledge into AI models.

The quantitative performance improvements achieved through Project Warp Speed demonstrate the AI’s capacity to exceed human performance benchmarks when armed with tribal knowledge — according to project partner and Anduril CIO Tom Bosco, the company has reported a 200x efficiency gain in their ability to anticipate and respond to supply shortages.

At Panasonic Energy’s Nevada Gigafactory, AI-powered quality control systems serve as a secondary judgment factor on production lines, scanning products and reducing waste by 10 to 15% compared to traditional manufacturing protocols, per an EV Magazine interview with Justin Herman, Panasonic Energy’s Vice President and Chief Information Officer.

Nguyen also provides another example where, instead of “picking up the phone to see if material shortage can be resolved,” AI agents constantly query warehouse management systems, compare 3PL info with ERP, and suggest solutions.

By grounding “smart” systems in actual user behavior, Nguyen’s vision for industrial AI offers manufacturing leaders a concrete path towards a model where institutional expertise and digital agility reinforce each other company-wide.

According to Nguyen, platforms don’t only aggregate real-time data on inventory flows, production cycles, or equipment status — they also encode the strategic decision-making frameworks that drive operational excellence.

By digitizing both the data and the reasoning processes behind efficiency gains, manufacturers can scale not just their outputs, but also the institutional intelligence that makes sustained competitive advantage possible across global operations.