Member of Technical Staff, ML Engineer 📍 Cambridge, MA · Hybrid, 3 days in office · Full time 🔬 THE OPPORTUNITY Join an early-stage team building AI systems to discover new physics at scale. Turn research code into reliable production systems and own the training and inference infrastructure behind AI-driven scientific discovery. ⚙️ WHAT YOU WILL OWN • Translate research code into dependable services with clear interfaces, meaningful tests and operational ownership. • Build services for launching experiments, routing across models and delivering results. • Own distributed ML jobs, GPU scheduling and model serving on a shared engineering platform. • Develop deep expertise in one area: distributed training, inference serving, ML platforms, or LLM gateway and agent infrastructure. 👤 WHAT YOU WILL BRING • 3+ years building and operating production ML training or inference infrastructure at meaningful scale, including ownership through failures. • 3+ years of software engineering on systems used by other engineers or researchers. Depth in at least one of: • Distributed training: PyTorch DDP/FSDP, Ray, multi-node jobs, checkpointing and performance. • Inference serving: vLLM, SGLang, Triton, batching, KV cache and quantization. • ML platforms: Kubernetes GPU scheduling, experiment tracking, registries and metering. • Gateways and agents: model routing, agent runtimes, sandboxes and evaluation infrastructure. 🔎 USEFUL EXPERIENCE • Debugging training loops, reward signals or inference behavior alongside researchers. • GPU infrastructure, CUDA or ML performance engineering. • GCP or AWS and infrastructure as code. • Experience turning research prototypes into production systems. 🤝 HOW THE TEAM WORKS High technical standards, simple designs and ownership from specification through deployment and on-call. Expect daily use of agentic coding tools and close collaboration with researchers. 💼 COMPENSATION Competitive salary, benefits and meaningful early-stage equity.