Senior Forward Deployed Engineer

• Most companies make you choose: build the platform, or go deploy it. Here you do both — and that’s the point
• As a Senior Forward Deployed Engineer, you’re part of a single team that both delivers Afresh’s AI into enterprise grocery customers and builds the platform that makes that delivery fast
• You’ll spend dedicated time in the field — embedded with a customer, integrating into their data, shipping AI systems on top of it — and dedicated time on the platform, turning what you just learned into reusable tooling the whole team deploys next. You build the house you live in
• Afresh leads the customer relationship and direction; you and a small team bring the technical firepower — scope and architect the work with the customer, then build it
• Because you also own the platform underneath, the rough edges you hit in the field become the things you fix at the root
• This is senior, hands-on, 0-to-1 work in a space with no playbook
• In the field (forward deployed):
• Partner with Afresh’s account lead and the customer’s technical teams to scope and architect the work — the data sources, the architecture, and the path to production
• Embed with the customer’s data and engineering teams (remote and on-site); integrate into their cloud and data platform; build production-grade pipelines and model messy enterprise data into trustworthy data products
• Design and ship LLM- and agent-powered systems on that data — retrieval, agentic workflows, data-quality and analytics agents — reliable enough to run in production, not just to demo
• On the platform (building the house you live in):
• Harden what works in the field into the shared platform: the knowledge and grounding layer (knowledge graph, ontology, and retrieval) that makes grocery data usable by LLMs, the agent frameworks, and the serving infrastructure
• Build the evals, tracing, and tooling that let the team measure quality — accuracy, hallucination rate, latency, cost — and ship faster on the next customer
• Build for leverage: clean interfaces and reusable building blocks, not one-off per-customer code
• Across both:
• Own the flywheel: field learnings flow straight into the platform, and platform improvements show up at the next customer

Benefits

• Comprehensive health plans
• Generous parental leave
• Equity packages
• 401k matching
• Flexible vacation policy
• Professional development program
• Work from home stipend

• Real data-engineering depth: building and operating data pipelines, modeling messy enterprise data, and working in a modern cloud data platform (Databricks, BigQuery, Snowflake, or similar)
• Genuine AI/LLM depth — you’ve built real systems with LLMs and agents (retrieval/RAG, tool-use) and you evaluate quality rather than eyeball it
• 3+ years building production software and data systems, with strong, production-grade code
• Customer-facing comfort: you work well with a customer’s engineers and data teams — running working sessions, explaining your thinking, and earning trust through what you deliver
• We encourage all highly-qualified candidates to apply, even if they do not fulfill all the listed criteria
• An architect’s instinct: you can take an ambiguous problem and a messy data landscape, design a clean and workable solution, and then build it
• Range across both modes — you genuinely like being in front of customers and going heads-down to build reusable infrastructure, and you can switch between them without one suffering. This is the role’s defining trait
• A bias toward ownership and momentum, and comfort traveling to customer sites regularly (~10-20%)
• Experience in grocery, retail, or supply chain data domains
• Knowledge graphs, ontologies, or semantic layers in production; graph and vector stores (pgvector, Pinecone, Weaviate) and hybrid search
• MCP or similar tool/context protocols; agent frameworks (e.g., LangGraph); MLOps, model serving, and observability for LLM systems
• Prior forward-deployed, solutions, or implementation engineering — or early-stage startup experience navigating rapid customer expansion