Founding ML Engineer, Post-Training
Core
Lead the transition of Newton, Aghanim's AI product, from proprietary model APIs to self-hosted models owned by the company.
Role type
Founding ML Engineer (Senior/Principal)
Builds
Self-hosted LLMs and agentic systems for game monetization and liveops automation
Domain
Gaming, Fintech, AI/ML
Deliverable
production ML models
Required skills
Post-training LLMs (SFT, LoRA/QLoRA, DPO, distillation), building evaluation systems from production data, agentic systems design, model serving (vLLM, quantization), Python, SQL, TypeScript
Preferred skills
Migrating production products from proprietary APIs to self-hosted models, RL for LLMs (GRPO), Langfuse/Braintrust experience, GCP (Vertex AI, Cloud Run GPU)
Technologies
Python, TypeScript, SQL, Langfuse, TRL, Unsloth, Axolotl, vLLM, GCP, Modal, Qwen, Gemma, DeepSeek, GLM
Responsibilities
Build production data pipelines for model training and evaluation; Develop benchmarks and LLM-judge evaluators; Compare and route tasks between frontier and open-weight models; Train and adapt models using distillation and RL; Deploy models and manage rollout based on quality benchmarks
Seniority
Senior/Principal, hands-on IC
