Machine Learning Engineer | Python | Pytorch | Distributed Training | Optimisation | GPU | Hybrid, San Jose, CA

About the role

Enigma is looking for an engineer to bridge the gap between research prototypes and production-ready services. You will focus on scaling model training and optimizing inference performance to ensure our systems remain cost-effective and reliable.

Key facts

What you'll do

• Transition research models into production services while meeting specific targets for latency, availability, and cost.
• Manage large-scale training jobs across multiple GPUs and nodes, focusing on throughput and training time reduction.
• Apply efficiency methods such as quantization, pruning, distillation, and Flash Attention to improve performance without sacrificing model quality.
• Develop and support model serving infrastructure using tools like vLLM, Triton, TGI, ONNX, TensorRT, and AITemplate.
• Connect models to data infrastructure including Parquet, Delta, and vector databases like FAISS, Milvus, Pinecone, or pgvector.
• Monitor performance metrics and capacity requirements to drive continuous system improvements.
• Coordinate with ML Ops and research teams to manage model registries, CI/CD pipelines, and evaluation workflows.

Requirements

• Bachelors degree in Computer Science, Electrical Engineering, Computer Engineering, or a related field.
• 3 to 5 years of professional experience in ML or AI engineering, specifically with production-scale training or serving.
• Proven track record of deploying high-throughput, low-latency ML services.
• Experience working across cross-functional teams including Research, Data, and Platform engineering.

Nice to have

• Masters degree in a relevant technical field.
• Practical experience with distributed training techniques such as DDP, FSDP, ZeRO, or tensor parallelism.

Skills & tools

• Primary language: Python.
• Frameworks: PyTorch, TensorFlow.
• Optimization: PTQ, QAT, AWQ, GPTQ, KV-cache optimization.
• Serving: vLLM, Triton, TGI, ONNX, TensorRT, AITemplate.
• Data: SQL, NoSQL, Parquet, Delta, FAISS, Milvus, Pinecone, pgvector.

Practical notes

• A Masters degree or equivalent industry experience is preferred over a Bachelors degree alone.