AI/ML Research Engineer
Work on foundational models, fine-tuning, and novel architectures. Experience with transformers, RLHF, and distributed training required.
About the Role
As an AI/ML Research Engineer at Hanzo, you will work on the core intelligence layer powering the Hanzo open cloud. You will train, fine-tune, and optimize frontier models, explore reinforcement learning with human & AI feedback (RLHF/RLAIF), and develop next-generation reasoning architectures.
Hanzo builds an open AI cloud — compilers, models, distributed infrastructure, and agent tooling. Our codebases are predominantly open source, and team members operate with high autonomy and end-to-end ownership.
Key Responsibilities
Train and fine-tune foundation models across text, code, multimodal, and agentic workflows.
Design and execute large-scale distributed training runs across GPU clusters using PyTorch, DeepSpeed, and Megatron.
Develop novel evaluation harnesses and benchmarks to assess model reasoning, tool use, and safety.
Collaborate with the inference and compiler teams to quantize, optimize, and deploy models to production.
Publish open source weights, research papers, and reproducible training recipes.
What We're Looking For
Deep understanding of transformer architectures, attention mechanisms, and modern LLM pre-training / post-training techniques.
Hands-on experience with PyTorch, distributed training frameworks (FSDP, DeepSpeed, Ray), and GPU performance profiling.
Demonstrated track record of shipping ML models to production or high-quality open-source / published research contributions.
Strong systems fundamentals: concurrency, networking, and memory efficiency.
Nice to Have
Experience with synthetic data generation, agentic reasoning loops (MCTS, ReAct), or tool-calling tuning.
Familiarity with Triton, CUDA kernel development, or custom model quantization formats.