SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Constructor powers product search and discovery for major retailers worldwide, processing billions of requests weekly. The company differentiates by optimizing for business metrics (revenue, conversion) rather than feature count, treating search as a machine learning challenge from the ground up.
You will lead two critical teams—Machine Learning & Recall and Query—sitting at the entry point of Constructor's discovery pipeline. This role bridges query understanding (intent, entities, language parsing) with candidate retrieval across massive product catalogs.
Key responsibilities include:
- Define multi-quarter strategy across both teams, balancing state-of-the-art techniques (vector search, semantic parsing, dense embeddings, LLM-assisted interpretation) with low-latency production execution.
- Lead the Query team in advancing tokenization, spell correction, entity extraction, intent classification, and multilingual query parsing.
- Advance the Recall team in evolving candidate generation architecture, combining keyword search with dense embeddings, vector retrieval, and hybrid recall models.
- Own measurable business outcomes: search-attributed conversion, revenue per visit, and GMV across retail verticals.
- Partner with ML researchers, data scientists, and engineers to move state-of-the-art models from research to cost-effective production systems.
- Make data-backed trade-off decisions balancing model complexity, accuracy, latency, and compute costs.
- Build robust offline and online evaluation frameworks connecting retrieval precision and query accuracy to customer revenue.
- Ensure clean handoffs between query parsing, candidate generation, and downstream ranking systems.
- Lead technical triage for query and retrieval anomalies, building durable automated fixes.
Success metrics in year one: measurable improvement in long-tail query precision, quantifiable lift in search-attributed conversion and revenue per visit, integrated evaluation framework, and optimized inference latency and compute infrastructure.
Constructor is Series B, remote-first, and growing ~2x year-over-year despite market headwinds. The team values empathy, openness, curiosity, and continuous improvement.
