General Intelligence Holdings
Founding Research Engineer — RL Environments & Post-Training
How your CV stacks up
Upload your CV to see how well it fits this job role
?%
General Intelligence Holdings · UK (remote-first, time on-site with our firms) · Founding equity
The one-paragraph pitch. We're a holding company that acquires UK accountancy firms and trains AI agents to run them — post-trained open-weight models, trained by reinforcement learning inside replayable replicas of the firms themselves, scored by deterministic verifiers built from work real accountants actually signed. Everyone else in this market rents access to someone else's operations or pays crowdworkers to grade work they couldn't sign. We own the firms, so we own the ground truth. You'd be the first engineering hire, building the machine that makes that sentence true.
What you'll build. The mirror pipeline: point-in-time, fully replayable replicas of acquired firms — ledgers, bank feeds, documents, correspondence — with sub-second deterministic resets. The environment harness: real engagements as RL tasks, agents acting through the same tools our staff use, traces hash-chained end to end. The verifier layer: porting our production evaluation gates (currently at a 0.00% false-auto rate across 804 live runs) into reward functions hard enough to survive gradient pressure. And the post-training loop itself: RL on open-weight models against those environments, on rented compute — our funded target is a specialized model that beats a prompted frontier model on our own gates, published either way.
Reasons to use Rodeo
Start with a chat, not a search bar
Grad scheme, placement, apprenticeship? Not sure what you want yet — that's fine. Your agent talks it through with you and turns "I have no idea" into a shortlist.
It searches the market for you
Every day your agent scans the market matching roles against what actually matters to you, not just keywords on a CV.
Only hits
No noise. No "maybe this fits." Just roles with a clear explanation of why they're right — and where to focus when applying.
You, probably. You've done post-training for real — RLVR, GRPO/PPO-family methods, or serious fine-tuning of open-weight models — at a lab, a training-platform startup, or as the person who built this alone because nobody else would. You've felt the unglamorous truths: that environment reset speed is a GPU-economics problem, that reward hacking is an adversary not an edge case, that eval contamination is a discipline not a footnote. You're pragmatic full-stack enough to own a Python harness, Docker sandboxes, and a Postgres data spine without a platform team. You work fluently with AI coding agents — this whole company is built that way. Accounting knowledge: none required. Respect for verifiable ground truth: non-negotiable.
What you get. Founding-engineer equity that means something. A dataset no lab can buy and no competitor can replicate without buying companies. The entire technical surface — you're not inheriting a codebase, you're setting its standards. And a working production system on day one: live gates, a live design-partner firm, real signed engagements waiting to become environments. Direct with the founder, no layers.
Get help with your application
Your very own career expert that helps elevate your application to the next level.
The honest stage disclosure. We're closing a pre-seed round; platform proofs are live in production at our design-partner firm and our first acquisition is in negotiation. You'd join as the round lands. If you need a big-company safety net this isn't it; if you want to be the person who built the first self-improving holding company, it is.
To apply. Skip the CV essay. Send: (1) the training loop, environment, or eval system you're proudest of and the ugliest problem you hit building it, (2) your take — three sentences max — on why verified reward beats LLM-judge reward for training agents on professional work, and (3) a link to anything you've built we can actually look at
“It took my CV and asked me questions relevant to understanding what kind of jobs to suggest for me. Suggestions were almost perfect. Jobs were exactly what I’ve been looking for.”
Jessica, London
Skills
Location