Afresh, the AI platform for grocery, began by tackling the most complex problem in the industry: fresh, and has evolved into the core AI platform for grocers. By leveraging proprietary AI designed for high-volatility environments, Afresh empowers partners like Albertsons, Meijer, and Wakefern to drive smarter decisions across their entire enterprise. Afresh has scaled to 6 enterprise-grade solutions, live in over 10% of the U.S. grocery market, and prevented over 200 million pounds of food waste last year alone.
About the Role:
As a Staff AI Platform Engineer, you build the AI and data platform that powers Afresh's products: the knowledge and retrieval layer that makes grocery data reliably usable by LLMs, the agent systems built on top of it, and the evaluation and serving infrastructure underneath. Your "customers" are Afresh's own engineers and AI products - your job is to give them a platform that turns raw grocery data into context a model can be trusted with, at production quality. This is senior, 0-to-1 platform work, making foundational choices about how the company represents grocery knowledge, grounds its models, and measures whether any of it is actually working, in a fast-moving space with no playbook.
Responsibilities:
• Design and operate the knowledge graph and ontology that capture how grocery data relates
• Build the retrieval systems (vector, graph, and structured) that feed the right context to models so grounding is reliable, not lucky
• Build LLM-powered agents (tool-use, multi-step reasoning, orchestration) and the serving infrastructure to run them reliably and cost-effectively
• Build the tools, abstractions, and interfaces other engineers depend on - a platform, not one-off features
• Stand up eval sets, LLM-as-judge harnesses, tracing, and observability, plus metrics (faithfulness, accuracy, hallucination rate, latency, cost) that show whether a change helped or hurt
• Build the pipelines, data products, and experimentation on Databricks/MLflow that take work from prototype to production, and partner with the engineers deploying AI in the field to harden what works into reusable capabilities
Requirements:
• 5+ years building production software, data, or ML systems; strong systems and API design skills in Python
• Hands-on production experience with LLM systems: retrieval/RAG, agents and tool-use, prompt and context engineering, and evaluation
• Solid data-engineering and data-platform foundations: pipelines, data modeling, and a modern cloud data stack (Databricks/Spark, MLflow, cloud warehouses)
• Comfort in the messy middle of AI systems - retrieval quality, latency and cost trade-offs, non-determinism - and the instinct to build guardrails and evals that make them trustworthy
• A platform mindset: builds for leverage and clean interfaces, and thrives in ambiguity in a fast-moving space
• Nice to have: knowledge graphs/ontologies/semantic layers in production; graph databases; vector stores (pgvector, Pinecone, Weaviate); MCP or agent frameworks (e.g., LangGraph); MLOps and model serving at scale; experience in grocery, retail, or other complex enterprise data domains
This is a hybrid role based in the San Francisco office (2 days/week onsite).
Salary Range: $168,912 - $253,368 + meaningful early-stage equity + benefits
Benefits: Comprehensive medical, dental, and vision coverage with the majority of premiums covered by Afresh, plus mental health support and counseling services. Competitive base salary, meaningful equity, and a 401(k) program with company match. Home office stipend and Coworking Wallets for flexible workspace access. Annual professional development budget. Monthly stipends for wellness/lifestyle and telecommunications. Flexible paid time off.
For more information or to apply, visit the Afresh careers site. Location: San Francisco, CA (Hybrid, 2 days/week onsite). Afresh provides equal employment opportunities to all employees and applicants for employment.
