Generative AI Team Lead

About the job

At KPMG Israel, our Generative AI delivery team builds end-to-end AI-powered products for clients across diverse industries. Our projects range from intelligent API services to complete production systems including user management, complex databases, process orchestration, and cloud infrastructure. We combine strong software engineering with AI capabilities to deliver solutions that create real business value.

About the Role

We're looking for a hands-on Generative AI Team Lead to both build and guide. You'll architect and ship production GenAI systems while mentoring a team of engineers - keeping one hand on the keyboard and one eye on the roadmap. This isn't a seat at a whiteboard; it's a role for someone who can write a RAG pipeline in the morning and run a technical design review in the afternoon.

What you'll do

Lead the design and delivery of end-to-end GenAI solutions, from prototyping to production - LLM-powered applications, agents, RAG systems, and fine-tuned models.

Stay hands-on: write production-quality code, review PRs, and set the technical bar for the team.

Define architecture patterns for prompt engineering, context management, retrieval pipelines, and model evaluation.

Mentor and grow a team of AI/ML engineers -run design reviews, pair-program, and provide technical feedback.

Partner with product, data, and platform teams to scope feasibility and drive AI initiatives from ideation to deployment.

Evaluate and introduce new models, tools, and frameworks (LangChain, LlamaIndex, vector DBs, etc.) as the space evolves

Requirements

3+ years of software engineering, with 2+ years in ML/AI roles

Hands-on experience building and shipping LLM-based applications (RAG, agents, fine-tuning)

Proficiency in Python; strong command of ML/AI libraries

Experience with at least one major LLM provider (OpenAI, Anthropic, Cohere, open-source)

Proven ability to lead or mentor engineers in a technical context

Comfort working in cloud environments (AWS, GCP, or Azure) with MLOps practices