Research Engineer
You will build the systems our research runs on, and run the experiments yourself.
You’ll work across training, post-training, data, evals, and agent backends, turning real company-building traces into models and systems that improve over time. This is a role for someone who can move from a paper to a training run to a production harness without waiting for a clean handoff.
What you’ll own:
- 01Post-training pipelines for EICO models: fine-tuning, SFT, RL, reward modeling, preference data, and evaluation loops.
- 02Datasets built from real company-building traces: demand discovery, customer acquisition, delivery, revenue, and verification.
- 03Eval harnesses that test whether agents can perform useful company-building work, not just benchmark tasks.
- 04Research tooling, model infrastructure, and agent harnesses used by the AI and incubation teams.
What we’re looking for:
- Exceptional engineering ability. Python required; PyTorch, JAX, or equivalent strongly preferred.
- Hands-on experience training, fine-tuning, post-training, or evaluating large models.
- Has built model infra, eval harnesses, agent systems, data pipelines, RAG, tool-use systems, or inference routing.
- Can read papers, form opinions, implement quickly, and tell whether the result actually matters.
- High standards for experiment quality, code quality, and speed.
- No PhD required, but strong research taste required.
- Grinder: a high-agency operator who removes bottlenecks instead of routing around them.
To apply, email careers@eico.so with a few lines about you and the most impressive thing you’ve built.