AI Context & Data Platform Engineer

Town; San Francisco (CA); Ibrido; USD 150.000 - 210.000; Tempo pieno

Descrizione del lavoro

Town is developing the most personalized AI assistant, starting from a strong identity model to power real work across tools. You will build the foundation of the search and data infrastructure to surface the right context—memory, messages, documents, and relationships—at the exact moment it is needed.

You will design a unified retrieval layer, balance lexical vs. semantic methods, and shape a durable data model that keeps content fresh and cost-effective at scale.

Competenze

• Experience with lexical and semantic search components across large-scale data systems.
• Ability to design and optimize search indexes and retrieval pipelines.
• Strong focus on latency, cost, and data freshness at scale.
• Systems thinker comfortable building foundational layers in greenfield projects.
• Willingness to take components from prototype to production.
• Bonus: knowledge graphs or retrieval for LLM/agentic systems.

Mansioni

• Build the search and retrieval layer providing context at the right moment for users.
• Combine lexical and semantic search and manage tradeoffs between vector and lexical approaches.
• Design durable data models to support fast, relevant context retrieval.
• Build and operate realtime/streaming and batch pipelines for data indexing.
• Stand up the indexing and storage layer to ensure fast, reliable access at scale.
• Lay groundwork for relating content by meaning across people, companies, and projects.

Conoscenze

Lexical and semantic search

BM25

Embeddings

ANN / vector indexes

Hybrid retrieval

Ranking

Descrizione del lavoro

Town is developing the most personalized AI assistant, starting from a strong identity model to power real work across tools. You will build the foundation of the search and data infrastructure to surface the right context—memory, messages, documents, and relationships—at the exact moment it is needed.

You will design a unified retrieval layer, balance lexical vs. semantic methods, and shape a durable data model that keeps content fresh and cost-effective at scale.