Modulus Labs is an AI systems engineering company. We design, build and operate production AI: the evals, guardrails, fallbacks and monitoring that keep a model useful long after launch day.
Most AI is a demo. We build the rest.
Modulus Labs exists because we kept seeing the same pattern. Talented engineering teams built AI prototypes that worked in the demo and collapsed in production. Not because the models were bad, but because the systems around them, the evals, the monitoring, the fallback logic and the security layers, were treated as afterthoughts.
We don't do AI research and we don't train foundation models. We build the production systems that turn what models can do into results a business can rely on, and we hold ourselves to the engineering standards we'd expect from any team we worked with.
Most AI consultancies optimize for the demo. We optimize for the six months after deployment, when models drift, data shifts, edge cases multiply and costs compound. That's when the real work begins, and it's the work we're built for.
Mujtaba Raza & Amaan Saigol, founders
Mujtaba leads the company: its strategy, its partnerships and its commercial direction. A lawyer by training, he moved into energy as the founder and CEO of Solar Citizen, a solar company, leading it from its first projects to real-time monitoring of solar operations. He brings that experience of running a business, along with legal reasoning and commercial judgement, to every engagement.
Amaan runs operations and delivery. He oversees the technical side of every engagement, from architecture to launch, manages each project day to day, and is the client's main point of contact throughout, so progress, decisions and trade-offs are always clear. He works with clients across the US, Europe and the Middle East.
We measure our work by what ships and what it achieves, not by hours logged. Every engagement has defined success criteria, and we hold ourselves to them.
We'll tell you when a simpler solution is better, and when AI isn't the right tool. Our job is to solve your problem, not to sell you more AI.
Clean code, considered architecture and real tests. Systems that are a pleasure to maintain, not just systems that work today.
You see our architecture decisions, our trade-offs, our progress and our mistakes. No black boxes, and no surprises at the end of a sprint.
We transfer knowledge as we build, through documentation, pairing and architecture walkthroughs. When we leave, your team can run and extend what we built.
Every engagement moves through the same four stages, whatever its size.
We map the workflow, the data and the risks, agree what success means in numbers, and build the evaluation set before any feature.
Small, reviewed increments against that evaluation set. Every change is scored before it merges.
Release pipelines with rollback, guardrails on every path in and out, and monitoring live from the first day.
We watch quality, latency, cost and drift in production, and keep improving the system after launch.
A defined system, built end to end against agreed success criteria, then handed over with its runbooks.
Our engineers inside your team, shipping alongside your people and leaving them stronger.
Architecture, reviews and build-or-buy decisions on a fractional basis, without a full build.
Every team claims quality. This is what ours looks like in practice.
We build the eval suite before the feature. Every AI component has measurable quality criteria, tested automatically on every change.
Every line is reviewed by a second engineer. No exceptions for prototypes and no shortcuts for urgent work.
Every system ships with logging, metrics and alerting on model quality, latency, cost and error rates, not just uptime.
An AI-specific security review on every project: prompt-injection testing, output validation, PII leak detection and access-control checks.
Architecture decision records, operational runbooks and API documentation ship with the code, because they're part of the product.
Each one is described in the terms the granting organization uses.
NVIDIA's program for startups building on accelerated computing. Admission requires an incorporated company,
a working product and engineering staff.
About the program (opens in a new tab)
Innovative AI Award
For clinical intelligence built for a premium diagnostics clinic: clinician-ready health narratives in under sixty seconds, and three times the patient follow-through.
Describe the problem, not the solution you have in mind. We'll tell you how we'd approach it, and whether AI is the right tool at all.