The opportunity
Join Instacart's mission to transform grocery shopping through frontier AI. As a Senior Software Engineer on AI Labs & Foundations, you will design, build, and operate the high-scale production systems that power our most ambitious AI experiences—from Cart Assistant, our…
What you'll do
Design, build, and operate production AI-powered systems and agentic: experiences (including Cart Assistant and voice AI) that directly impact how millions of customers shop.
Build foundational systems for cutting-edge AI experiences, ranging from: embedding infrastructure and voice AI pipelines, to client facing components and integrations, by prototyping bold ideas and productizing what works.
Integrate foundation models via APIs and open-source frameworks; apply: techniques like retrieval-augmented generation and vector search where appropriate.
Own projects end-to-end: requirements, technical design, implementation, testing, deployment, observability, and iterative improvement focused on reliability, latency, and cost efficiency.
Collaborate with cross-functional partners in product, design, data science,: and infrastructure to ship AI features end-to-end.
Drive engineering excellence, including thoughtful system design, rigorous: code review, and technical leadership that includes defining and promoting best practices for AI/ML production engineering across the team.
What they're looking for
- Proven senior software engineer who has built, shipped, and operated: production systems at scale. You make architectural calls, own what you build, and deliver through ambiguity.
- Hands-on experience with AI or ML in production. You've shipped LLM-powered: features or integrated foundation model APIs into a live product, demonstrating the necessary expertise at the intersection of robust software engineering and deep production ML.
- Experience owning services end-to-end, including CI/CD, automated testing,: observability (logging, metrics, tracing), and on-call participation.
- Strong communicator who partners well across disciplines: you want to get to the right answer, not just defend the first one.
