Software Engineer, Knowledge Engine – Query Engine

Description

The company is an applied AI lab building a search engine. It crawls the full web, trains embedding models to process what it finds, and runs vector databases to retrieve results. Its search already serves Cursor, Cognition, HubSpot and more than 400,000 developers. It has raised $350m from Lightspeed, Benchmark and a16z.

The long-term aim is search over all of the world's information, well beyond Google. The work centres on large-scale machine learning systems that will shape how AI systems take in information.

The Knowledge Platform team builds the company's knowledge engine. It continuously pulls facts from the web and other sources into one of the largest open-domain knowledge graphs, which holds hundreds of billions of evidence-backed claims. The team splits its work between construction pipelines that judge what to trust and retrieval engines that decide how to search the graph. The goal is to let agents answer any factual question about the world with very high precision and completeness.

This role owns the query engine. It searches hundreds of billions of claims spread across several retrieval backends, including a distributed graph database and vector, full-text, time-series and geospatial indexes.

The work is greenfield with a large scope. No existing system answers precise questions over hundreds of billions of evidence-backed claims at agent speed, so you will make foundational calls on storage layout, indexing across engines, the query language and the planner that routes work between them. Every query depends on those choices.

Responsibilities

• Design the data layout, partitioning and indexes for graph, vector, full-text, time-series and geospatial retrieval, so that bulk snapshot builds and live ingest never bring the system down.
• Build cost-based planning that uses cardinality estimates to pick the backends for each question, their order and how their results are joined, running traversals, retrievals, filters and aggregations as one plan.
• Serve thousands of queries per second with low p50 and p99 latency, controlling memory and cost per query under heavy concurrent load, from on-disk layout through to network fan-out.
• Control what a query can see while the data converges. The graph takes live updates and bulk snapshots at the same time and is eventually consistent by design, so you will handle snapshot promotion, the order live updates are applied and stable pagination across versions.
• Measure precision and completeness with end-to-end checks that every deployment must pass, plus evals and metrics that reveal when a plan returns wrong results or misses results it should have found.

Requirements

• You have worked deep inside a distributed database, query engine, search engine or storage system.
• You have a strong understanding of indexing, query planning and optimisation, and storage formats.
• You have a solid grounding in distributed systems: consensus, replication, sharding and fault tolerance.
• You are strong in a systems language such as Go, Rust or C++.
• You profile before optimising and have shipped measured latency improvements to production.
• You care about correctness, reliability and data quality as much as scale.
• Bonus: experience with graph databases or with combining vector, full-text, time-series and geospatial retrieval.
• Bonus: experience with compilers, query language design or cost-based optimisers.
• Example projects you might take on include a system that retrieves and reasons over everything known about any company or person, answering factual questions precisely and completely.
• Example projects include answering precise multi-hop questions over hundreds of billions of claims at thousands of QPS with interactive latency.
• Example projects include a planner that picks the right mix of graph, vector, full-text and geospatial retrieval for any question and shows that it chose well.
• Example projects include making new knowledge queryable seconds after it is synthesised, without ever returning an inconsistent result.

Conditions

The company will sponsor international candidates for visas, including STEM OPT, OPT, H1B, O1 and E3. It cannot guarantee any visa, but it has a strong record of sponsoring candidates from around the world, and after an offer the team will work hard to secure one.

Benefits

• Premium medical, dental and vision healthcare.
• Fertility benefits.
• 16 weeks of fully paid parental leave for all new parents.
• A monthly wellness stipend for all employees.