Engineering & Technology - Engineering
Software Engineer - Search & Retrieval
IC / Practitioner
Independently auditedBuilds and operates the software the business runs on. Owns a defined area of the function and the outcomes it produces.
Audited package
Built, independently audited role package. Everything on this page is written for this seat.
Your Stack (Swappable)
This role works against whatever tools you already run. We list the tool categories the role operates in, never a named vendor, so you are never asked to replace a system that works.
- Search index
- Design & Prototyping
- Query and ranking services
- Analytics & BI
- AI / LLM Platform
- Experimentation & Feature Flags
- Observability & Monitoring
- Data Warehouse & Lakehouse
Deliverables
- Design documents
- Review checklists
- Technical standards
- Delivery and sequencing plans
Responsibilities
- Build document ingestion, normalization, tokenization, metadata, indexing, deletion, update, and freshness pipelines.
- Implement lexical, semantic, vector, hybrid, graph, or domain-specific candidate retrieval with measurable recall.
- Build query parsing, intent classification, rewriting, spelling, synonym, entity, and filter behavior.
- Integrate ranking features and models, feature retrieval, score computation, reranking, diversity, and business constraints.
- Define and operate offline evaluation using judged query sets, NDCG, precision, recall, MRR, zero-result, and slice metrics.
- Run online experiments with click, success, reformulation, abandonment, latency, safety, and access-control guardrails.
- Maintain index versioning, freshness SLOs, backfill, deletion, rollback, and corruption recovery.
- Instrument query traces, candidate loss, feature errors, rank divergence, tail latency, and result explanations.
The essence of this seat
The Software Engineer - Search & Retrieval is the engineer who builds the production search stack from document ingestion and indexing through query understanding, candidate retrieval, ranking, and result delivery.
Signature move
Trace one query end to end. Capture query text, parse, rewrite, filters, index version, candidate stages, feature values, rank scores, and final results before proposing a fix
What good looks like
separates recall from ordering before tuning the ranker
Decision rights
- - - The implementation and operation of document ingestion, index construction, query understanding, candidate retrieval, filtering, ranking integration, and result-serving behavior
Working vocabulary
What this role does
The Software Engineer - Search & Retrieval is the engineer who builds the production search stack from document ingestion and indexing through query understanding, candidate retrieval, ranking, and result delivery. The role's domain is lexical and vector retrieval, index construction and freshness, query parsing and rewriting, candidate generation, ranking features, relevance evaluation, filtering, tail latency, and online search quality. Its core outcome is a user query returns fresh, permitted, diverse, and relevant results within a predictable latency budget, with an evaluation trail that explains regressions and tradeoffs. This is a standalone 1:1 role package. It represents only the Software Engineer - Search & Retrieval title. A generic Software Engineer package does not replace it, and this package does not replace any neighboring domain, layer, platform, or level title. The differentiation must remain visible in the role's metrics, artifacts, failure modes, operating protocols, stakeholders, interview loop, and Golden Q&A. The altitude and seams are explicit. The Software Engineer owns general product delivery without this retrieval mandate, while this seat owns the search and retrieval domain, relevance evidence, indexing, ranking, and query behavior. The Software Engineer - Applied Inference owns efficient production inference for the model used inside the path, while this seat owns the end-to-end retrieval and ranking product path. The Data Scientist / Relevance Scientist owns offline analysis, experimental methodology, or model research rather than the serving system, while this seat builds and operates the production retrieval system and online behavior. Two mental models should be visible in every decision: Candidate recall sets the ceiling and Offline relevance is a proxy. The work is strong only when those models appear in the evidence rather than remaining vocabulary.
Best-fit customer
engineering teams that want consistent judgement applied to design, review, and delivery
Need more than one role?
Buying the whole team or the complete department lowers your effective cost per role at every tier of the ladder.