Research Engineer
Build benchmarks and backtests, train models, and test whether our agents make sound financial decisions.
About the role
As AI moves from doing financial work to informing investment decisions, the hard problem is knowing when its judgment can be trusted and how to improve it. We are looking for a research engineer who can measure that judgment rigorously and turn the results into better models and agents.
You will work closely with AI researchers, engineers, and investment professionals across the research loop: evaluating the agents in our products, testing forecasts and investment decisions against real outcomes, and training our in-house models. You will own experiments from design to published result, and much of what you build will become infrastructure the rest of the team relies on.
What you’ll do
- Design benchmarks for long-horizon, tool-using agents, with tasks drawn from real investment workflows, graders that agents cannot game, and results reported across repeated trials.
- Build backtests and live forecasting experiments that compare agents with professional consensus and fair baselines, and guard against look-ahead leakage and memorized market data.
- Post-train our own models with distillation and reinforcement learning, and build the environments, task sets, and rubric-based rewards that train them on agentic financial work.
- Build and improve agents, skills, and multi-agent systems for investment research, read their traces to diagnose failures and regressions, and separate real mistakes from flaws in the benchmark.
- Make training, inference, and large-scale agent experiments faster and cheaper, and check that each optimization leaves measured performance intact.
What you’ll bring
- Hands-on experience with large language models, such as post-training, building agents and their environments, or running evaluations at scale.
- Strong Python and research engineering skills, with a record of building training, inference, or evaluation systems that are reproducible and use compute efficiently.
- Rigorous experimental judgment: you hold out test data, size experiments for statistical power, account for variance between runs, and spot gains that come from leakage, contamination, or noise.
- Clear research writing and the habit of stress-testing your own results before you share them.
- Curiosity about how investment professionals make decisions, and the ability to own open-ended research problems with little direction.
Why Fundamental Research Labs
Fundamental Research Labs is building digital humans, starting with AIs that understand the fundamental value of a business. We bring research, engineering, and investment expertise together and turn what we learn into products people can use. Your work will help shape both the research and the products built on it.
Apply
Send your CV or LinkedIn profile and a short note about your interest.
Tell us about an experiment you ran on a model or agent, how you made sure the result was real, and what you changed because of it.
Apply by email