About the Company

Grey Machines is a decision intelligence platform: it lets you simulate a strategic decision before you make it. You ask a "what if" question in plain language, and the system builds a working model of that world from scratch, with its actors, feedback loops and delays, grounded in real data and cited evidence. It then plays out thousands of possible futures and shows the range of outcomes, what drives them, and a brief in which every number traces back to its source, in hours rather than the weeks a strategy team would take. It is built for strategy teams, investors and policy makers.

The Problem

Take a question like "What happens to nickel markets if Indonesia bans ore exports?" or "What does a US-Iran escalation do to oil and LNG?" Answering it well usually takes weeks: a strategy team, a financial model, expert calls and a deck.

Grey Machines answers it in hours. From one plain-language question, the system builds a model of that world from scratch: the actors, the stocks and flows, the feedback loops and the delays. It grounds the starting values in real data and cited evidence, then runs thousands of Monte Carlo futures, with convergence checks deciding when to stop. What comes back is an interactive causal loop diagram, the range of outcomes, and a written brief in which every number traces back to where it came from.

It works end to end and is in private beta. Now it has to be correct, fast and reliable at scale. That is the job.

What You Will Build

You will be the first engineer and the most senior technical person in the room, so the scope is all of engineering. You will make the architecture calls every later engineer builds on, and own the consequences.

- The simulation core: the ODE solver, the Monte Carlo engine, sensitivity analysis, and the convergence and health checks that decide when a result can be trusted

- The model-building pipeline: frontier LLMs build the model in stages, and deterministic validators and repair loops check every step. The real work is deciding what the LLM can be trusted with, what code has to verify, and how to fail safely.

- The data and evidence layer: values from sources such as FRED, the academic literature, the web and the documents users bring, with units, ranges and provenance that survive end to end

- The execution layer: long Monte Carlo jobs are checkpointed and resume after a restart, today on one machine. Next they need to be scheduled and run in parallel across many.

- Production: AWS, deploys, monitoring, tracing and incident response

- The frontend when it needs you: a React 19 app with Canvas 2D and D3 visualisations that make dense quantitative output readable for a strategist, not just a quant

- The engineering bar: testing, review and release discipline, and hiring the team that comes after you

Who You Are

- You have built systems from zero, not just shipped features into existing ones, and lived with your own architecture decisions, including the wrong ones.

- You are fluent in Bayesian probability and statistics, stochastic processes and numerical methods: ODEs, Monte Carlo and convergence diagnostics. A PhD is not required. Being able to reason about what the numbers mean is.

- You write production code fast, debug carefully and care when it is wrong. Production experience with Python, FastAPI, PostgreSQL and React is required: the codebase is built on them.

- You have run LLMs in production: structured output, prompt caching, multi-stage pipelines and the failure modes that only show up there.

- You write clearly. We work remotely and asynchronously, so your design docs and code reviews are how you work.

Strong Signals

None of these is required. They are backgrounds that tend to fit.

- A year or two at a quantitative trading firm (Jane Street, Hudson River Trading, Tower Research, Graviton or similar)

- Your own startup, including one that did not work out. We care about what you learned.

- A maths, CS, physics or statistics degree from a school these firms hire from

- Multi-tenant SaaS or high-throughput compute pipelines

- CTO, Staff or Principal Engineer experience at an early-stage company

What It Is Honestly Like

We are an early-stage startup, and we would rather say so up front.

- Priorities move. Something you build this month may be thrown away next month because we learned something.

- There is no team behind you yet. You will check your own blind spots and make calls with incomplete information.

- Some weeks will feel like a drawdown. We want someone who stays steady through the good weeks and the bad ones, and treats uncertainty as part of the job.

In return you get real ownership, founder-level say in the product and the company, and a genuinely hard applied-maths problem. Fully remote, flexible hours.