Role: Machine Learning Engineer

Design, train, and ship models that guide reviewers to the most important changes, resolve merge conflicts more safely, and make diffs understandable through semantic grouping and summarization. You'll combine retrieval, code‑aware modeling, and small‑model techniques to deliver low‑latency, cost‑effective ML in production.

Responsibilities

• -Train and fine‑tune models that predict the next location a reviewer should focus on: ranking files, hunks, and comments
• -Build ML‑assisted merge conflict resolution: align and reconcile hunks using code context/AST, propose safe resolutions, and explain changes
• -Develop semantic change grouping and diff visualization: detect refactors/moves, cluster related edits across files, and generate concise summaries
• -Evaluate retrieval and RAG for pull requests: retrieve relevant context, discussions, and history to improve grouping, prioritization, and summarization
• -Design offline/online evaluation: labeled datasets, ranking metrics (e.g., MAP/NDCG), A/B experiments, and human‑in‑the‑loop feedback
• -Operate efficient inference: apply small‑model techniques (LoRA, supervised fine‑tuning, distillation, quantization)

Qualifications

• -Strong background in deep learning research and implementation, with experience in modern architectures and concepts (transformers, MoE, reasoning, etc.)
• -Track record publishing results at top-tier conferences (NeurIPS, ICML, ICLR)
• -Expertise in Python and ML frameworks (PyTorch, JAX)
• -Experience with MLOps and model optimization (ONNX, TensorRT, serving infrastructure)
• -Familiarity with cost‑effective techniques: LoRA, supervised fine‑tuning, distillation, and quantization for small/efficient models
• -Familiarity with systems programming across the stack: Rust/C/C++, CUDA/GPU optimization, and kernel-level performance tuning
• -Experience with reinforcement learning, particularly in systems optimization and decision-making tasks