This is a Permanent Role with a Valued Clients of Hireahead.com
Lead a small, high-velocity team to design, build, and ship ML and GenAI products end-to-end, from problem framing and modeling to scalable deployment and monitoring.
Responsibilities:
Own the technical roadmap for ML/GenAI initiatives and align it with product goals and delivery timelines.
Lead LLM workflow and agentic system design, including retrieval, tool-use, and evaluation pipelines.
Guide model development using PyTorch and HuggingFace; enforce code quality, testing, and model evaluation standards.
Oversee MLOps: CI/CD for models, API design, containerization, orchestration, and cost/performance optimization on AWS.
Collaborate with product, data, and infra teams; run experiments, A/B tests and error analysis.
Mentor engineers, conduct code/model reviews, and grow team capabilities.
Must-Have Skills
At least 4 years of experience with a minimum of one year in technical leadership leading projects or teams.
General AI/ML Team Lead requirements: roadmap ownership, sprint planning, stakeholder communication, mentoring, and setting engineering best practices.
Strong Python understanding: idiomatic code, testing, packaging, async I/O, performance profiling.
Strong ML, DL, and GenAI understanding: supervised/unsupervised learning, transformer architectures, prompt engineering, evaluation, and safety/guardrails.
Experience building LLM-based workflows and agents, including retrieval (RAG), tools, memory, and evaluation loops.
Hands-on experience with LangChain, HuggingFace, PyTorch, Neo4j, Milvus, and AWS for scalable, production-grade systems.
MLOps experience: model deployments, API development, observability, infra management (containers, orchestration, IaC), and data/version governance.
Preferred Skills:
Knowledge graph design and Neo4j schema modeling for semantic retrieval and reasoning.
LLM fine-tuning with deployment to production: LoRA/QLoRA, PEFT, quantization, evaluation, rollback strategies, and cost optimization.
Startup experience: comfort with ambiguity, rapid iteration, and owning outcomes across the stack.