Manufacturing limited only by physics
US manufacturing is hitting its operational limits, and nowhere is that clearer than in data center construction, where AI compute demand is outpacing the physical infrastructure that supports it. Morgan Stanley forecasts $2.9 trillion in spending on data centers and related infrastructure by the end of 2028 and the factories producing chips, servers, racks, and power systems can’t keep up. They sit on massive latent capacity but system fragmentation within each facility prevent them from scaling.
Each factory is a custom distributed system: decades-old ERPs, ad-hoc Excel pipelines, siloed operational domains operating on disconnected quality, work-order, tooling, scheduling, and financial systems*.* That fragmentation means greedy decisions get made, and the factories are stuck in naive local minima because no one has a unified view. Without a structured, profit-aware model of what’s actually happening across production, inventory, quality, and dependencies, factories can’t prioritize the highest-leverage actions or respond in real time to shifting constraints.
We automate the painfully bespoke plumbing so factories can reason holistically. Large language models generate custom integration code for each facility’s unique stack - legacy APIs, undocumented exports, disparate telemetry, and tooling - stitching it into a queryable data fabric. On top of that data we run real-time analytics, simulations, and optimization logic to compute the true profit impact of possible actions. The highest-leverage recommendations are delivered directly into operators’ workflows via plain-English interaction.
Complement’s platform is fundamentally agnostic to industry, but we’re starting where the pressure is highest: the data center supply chain. At scale, we become the machine that builds the machine - bringing to life a cybernetic, self-improving manufacturing network.
James led the Guidance Navigation and Control flight software team at Relativity Space, designing and writing the software that controlled Terran 1 during its first launch. At Neuralink he developed systems for brain machine interfacing, including on-device implant control and machine learning models. He researched AI for treatment of Parkinson's disease at Cornell University and graduated summa cum laude in Computer Science and Neurobiology.
Alex was an Applied Scientist in Microsoft's Azure AI group. During his tenure he made significant contributions to cutting edge technologies such as personalized noise suppression and speech-to-speech translation. Alex holds a BSE in Electrical Engineering with a focus on robotics and cyber-physical systems from Princeton University, where he also led the Formula Hybrid team.