Python Developer – LLM Post-Training (Remote | San Francisco)
First seen by Alion on Sep 6, 2026.
We are a San Francisco-based AI infrastructure company working with leading frontier AI labs to build post-training data and evaluation infrastructure for foundation models. We are hiring a Python Developer to create high-quality datasets, reinforcement learning environments, and benchmarking pipelines used to improve and evaluate state-of-the-art LLMs. This is a remote role with flexible working hours.
Responsibilities
* Create and curate datasets for LLM post-training (SFT, RLHF, RL, preference optimization).
* Build and maintain RL environments for agent evaluation.
* Develop Python tooling for dataset generation, validation, and transformation.
* Evaluate models on custom benchmarks and testing pipelines.
* Collaborate with research and engineering teams to deliver client-specific post-training datasets.
* Work with terminal-first development workflows and cloud infrastructure.
Required Skills
* Strong Python programming skills.
* Understanding of LLM fundamentals and post-training concepts (SFT, RLHF, RL).
* Experience working with structured data (JSON, CSV, YAML).
* Git, Linux/Unix command line, and solid software engineering fundamentals.
Good to Have
Experience with RAG, agentic AI systems, Hugging Face Transformers, LoRA/PEFT, LangChain or LlamaIndex, vector databases (FAISS, Qdrant, Milvus, Pinecone, Weaviate, ChromaDB), Docker, AWS/GCP, FastAPI/Flask, Bash, CLI tooling, model evaluation frameworks, benchmarking, and AI infrastructure.
Compensation
Base Salary: USD $1,250/month
Equity: ESOP/Equity package included.
Performance Bonuses: Up to USD $4,000/month (in addition to base salary).
Location
Remote (Worldwide)
Work Hours
Flexible, remote-first, asynchronous work environment.
How to Apply
Apply here: https://tally.so/r/wLReJG
Please complete the application form and submit the required details. Only shortlisted candidates will be contacted.
Skills
Python, Large Language Models (LLM), Fine-tuning LLMs, PEFT (Parameter-Efficient Fine-Tuning), Amazon Web Services (AWS), Google Cloud Platform (GCP), Docker, Retrieval Augmented Generation (RAG), Vector database, FastAPI, Agentic AI, Hugging Face Transformers, LoRA / QLoRA, LangChain, LangGraph, LlamaIndex, Bash