Join a high-intensity ML team owning the post-training of Lumen Enterprise, the LLMs powering the world’s best autonomous coding agent. You will drive SFT, RL, and continued pretraining to push state-of-the-art performance on complex software engineering tasks. This role offers direct impact on a product used by global enterprises.
This job is no longer actively hiring. Talk to Jack to find live roles.
ML Engineer at Cosine.sh
Join Cosine.sh in London to lead the ML training of the world's most advanced autonomous coding agents. As an ML Engineer, you will own the SFT and RL pipelines for the Lumen Enterprise models, moving beyond simple autocomplete to build agents that reason through entire codebases. Working on multi-node GPU clusters with direct access to the CEO, your work will define the state-of-the-art in AI-driven software engineering. If you are a PyTorch expert who wants to see your models ship to enterprise customers instead of just publishing papers, this is the role for you.
Want to apply for this role?
This role is no longer actively hiring, but Jack can still help you discover similar open roles that fit.
Location
London, United Kingdom
Compensation
£80k-£110k + Equity
Company
Cosine
Role overview
Software
Cosine is an autonomous, on-premise coding agent post-trained on human reasoning data to deliver unmatched software-engineering accuracy, security, and speed for regulated enterprises.
What you will do
- Transform open-source base models into high-performance SWE agents through supervised fine-tuning and advanced reinforcement learning (PPO, GRPO, or DPO).
- Design and execute large-scale training experiments on multi-node clusters, optimizing for long-context stability and tool-use reasoning.
- Build and iterate on automated RL loops where models are rewarded for successfully running tests, linters, and static analysis on real-world codebases.
Who this is a fit for
- 3-5+ years of experience training deep learning models in production with deep proficiency in PyTorch distributed primitives like FSDP and DDP.
- Proven track record of training large-scale models (≥70B parameters) and implementing complex RLVR systems for LLM alignment.
- Strong software engineering background with the ability to write production-grade Python and a focus on data quality and sampling strategies.
Why this role is remarkable
- Direct ownership of post-training for Genie, a SOTA coding agent that achieved a 72% score on OpenAI’s SWE-Lancer benchmark.
- Work at the technical frontier with multi-node GPU clusters, large-scale MoE architectures, and long-context training on proprietary software-engineering reasoning data.
- Join a small, elite 4-person ML team reporting directly to the CEO, where your training runs ship immediately to real-world enterprise users.
How Jack & Jill work together
I get to know what you’re great at, then find roles you’d never find yourself.Ok, I'll go first. I'm Jack, an AI that gets to know you on a quick call, learning what you're great at and what you want from your career. Then I help you land your dream job by finding unmissable opportunities as they come up, supporting you with applications, interview prep, and moral support.
I recruit from Jack’s network and make the intro when I spot a great match.And I'm Jill, an AI Recruiter who talks to companies to understand who they're looking to hire. Then I recruit from Jack's network, making an introduction when I spot an excellent candidate.
Jack gets to know what you're great at and what you want next, then searches 15 million jobs daily and helps you discover roles at companies like this.
Meet Jack
What happens next?
Jack’s an AI agent for job searching and career coaching. He works for you.
Jill is the AI recruiter working for the company. She recruits from Jack’s network.
If your profile’s a match and Cosine wants to meet, Jill will make the intro. In the meantime, Jack will send you excellent alternatives.