> Markdown version of [/jobs/ext/4471631-ai-context-data-infrastructure-engineer](https://www.wearedevelopers.com/jobs/ext/4471631-ai-context-data-infrastructure-engineer). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# AI Context & Data Infrastructure Engineer

- **Company:** TWP Inc.
- **Location:** San Francisco, CA, United States
- **Contract:** Permanent contract
- **Skills:** Artificial Intelligence, Data Infrastructure, Data Recovery, Graph Database, Rapid Prototyping Process, Search Technologies, Indexer, Juniper, Virtual Agents
- **Published:** October 26, 2026
- **Apply:** https://www.careerbuilder.com/job-details/ai-context-data-infrastructure-engineer-san-francisco-ca--f5eadbad-3355-4238-b252-f13b70520a42

## About the Role

Artificial Intelligence (AI), Artificial Intelligence (AI), Customer Relationship Management (CRM), Juniper Networks M-Series, Data Recovery, Semantic Search, Data Modeling, Order Picking/Packing, Rapid Prototyping

## Description

Town is building the most personalized, most capable AI assistant for everyone - and personalization at that level is a retrieval and data problem. The assistant is only as good as the context it can bring into the moment: the right memory, message, document, or relationship, pulled fast and related by meaning across everything a person and their team touch.

You'll build the foundation the whole product reasons over: the search and data infrastructure behind that context. One shared retrieval layer combining lexical and semantic search, the realtime and batch pipelines that keep it fresh and correct, and the durable data model everything else is built on.

This is greenfield and high-leverage: you'll be the first person building this layer.

What you'll do

 * Build the search and retrieval layer that puts the right context at every Townie's fingertips, the moment it's needed - one shared layer the whole product pulls from instead of refetching context on its own.
 * Combine lexical and semantic search and own the tradeoffs between them: vector vs. lexical, precompute vs. fetch, hybrid retrieval, and ranking.
 * Design the durable data model the assistant's work is built on, so context is relevant, fast, and cost-effective.
 * Build and operate the pipelines behind it - realtime/streaming and batch - that keep the index fresh and correct as the underlying data changes.
 * Stand up the indexing and storage layer and keep it fast and reliable at scale: latency, cost, freshness, and completeness.
 * Lay the groundwork for relating content by meaning across everything the assistant knows - the start of a knowledge graph of people, companies, projects, and how they connect.