Graph Retrieval Engineer- Startup Backed Anthropic and Stripe

Join an elite, venture-backed team building next-generation, AI-powered collaboration tools for the enterprise.

Technical Integrity has been retained to lead the search for a rapidly growing startup developing AI systems that improve how large organizations think, communicate, and make decisions. Their product uses cutting-edge AI to identify and resolve coordination gaps automatically — helping teams operate more intelligently and efficiently at scale.

Following a recent and substantial round of funding, the company is expanding its world-class engineering team in Colorado and beyond.

We're hiring a senior staff engineer specializing in knowledge graph retrieval to solve a critical scaling challenge: our AI agents build massive English-language knowledge graphs of enterprise operations, and we need intelligent retrieval systems to extract relevant information from graphs that are 100-10000x larger than any LLM context window. This is a novel information retrieval problem combining classical search/ranking techniques with cutting-edge agentic LLM approaches, applied to highly unstructured natural language knowledge graphs.

Our product builds personal AI agents for leaders and managers in large, complex organizations. Each agent constructs a "world model" - an English-language knowledge graph capturing:

• Company goals and project hierarchies
• Cross-team dependencies and relationships
• Project status, risks, blockers, and opportunities
• People, roles, and communication patterns
• Decisions, commitments, and timelines

When a leader asks "what could cause Project X to run behind?" we need to intelligently traverse the knowledge graph to find:

• Upstream dependencies (projects Project X depends on)
• Status of those dependencies (are they at risk?)
• People involved (are they over committed?)
• Recent decisions that might impact timelines
• Communication patterns (are teams coordinating effectively?)

This isn't keyword search. This is graph-structured retrieval over unstructured natural language content where understanding business semantics (dependencies, criticality, risk) is essential.

• Hybrid Structure:
  Graph structure (dependencies, hierarchies) with unstructured content (natural language descriptions, status updates, Slack messages)
• Semantic Complexity:
  Need to understand business concepts like "critical path", "blocking issues", "resource contention", "scope creep"
• Agentic Retrieval:
  Classical techniques (graph traversal, embeddings, ranking) PLUS novel approaches (LLM agents crawling their own knowledge graphs)
• Dynamic Graphs:
  Knowledge graphs update continuously as we ingest chat messages, docs, emails
• Query Diversity:
  From precise ("who owns Project X?") to exploratory ("why might this initiative fail?")

This is a small team (11 people) building for massive enterprises. The ideal candidate needs to be able to handle the algorithmic core of this problem plus the surrounding tooling and infra:

• Designs and implements core retrieval algorithms
• Builds production infrastructure (caching, indexing, APIs)
• Instruments and optimizes performance
• Works directly with product to understand use cases
• Ships fast, iterates based on real user feedback

• Algorithms and data structures (graph algorithms especially)
• Complexity analysis and optimization
• Data systems architecture

• Search Ranking & Relevance: Should be able to discuss specific algorithms (TF-IDF, BM25, learning-to-rank models like LambdaMART) and explain when to use each.
• Evaluation & Metrics: Must understand precision, recall, F1, NDCG, MRR (Mean Reciprocal Rank). Should have run offline evaluations and A/B tests to measure retrieval quality.
• Knowledge Graphs
  - Implementation Depth: Should have built graph data structures or worked with graph databases (Neo4j, Amazon Neptune, GraphQL engines), not just queried them.
• Vector Search
  - Built Not Just Used: Should understand how vector indexes work (HNSW, IVF, product quantization), not just called Pinecone APIs.
• Hybrid Retrieval: Should have combined multiple retrieval signals (keywords + vectors + graph structure, or dense + sparse retrieval).
• Query Understanding: Should understand NLP techniques for parsing user intent—entity extraction, query expansion, semantic parsing, or intent classification.
• Multi-Hop Reasoning: Bonus if they've built systems that retrieve information across multiple…