Principal AI Engineer

Recruits Lab; Washington, Baltimore (District of Columbia, MD); Remote; USD 150,000 - 190,000; Full time

Job summary

Recruits Lab seeks an experienced AI systems engineer to build production NLP/IR pipelines for life sciences talent. You will own retrieval, scoring, and guardrails for high-stakes staffing, collaborating with leadership to ship measurable impact.

You will translate complex domain rules, drive end-to-end ML workflows, and ensure auditable outputs that support rapid, decision-ready candidate shortlists across regions.

Qualifications

• Shipped production NLP/IR or ranking systems evaluated by human judgment.
• Owns LLM applications with retrieval, structured extraction, evaluation, and guardrails for high‑stakes, PII‑sensitive workflows.
• Translates messy domain rules (territory, quota, multi‑stakeholder cycles) into features, labels, and product behavior recruiters will trust.
• Proven delivery as a principal IC: architecture, code, evals, and stakeholder alignment without a large platform org underneath you.
• Clear written communication that matches how we work: specific, outcome‑based, no jargon theater.

Responsibilities

• Ship, within 90 days, a production retrieval and ranking pipeline that improves passive-candidate shortlist quality (precision at 10) by at least 25% versus current sourcing on two live life sciences searches.
• Design and own an evaluation harness that measures match quality against hiring-manager calibration notes, quota/territory constraints, and named-account verification—not keyword overlap—and report lift monthly.
• Reduce time-to-calibrated-shortlist on repeat search types by 30% within six months by automating mapping of competitor talent pools and adjacent buyer relationships.
• Put LLM-assisted screening into the recruiter workflow so structured signals (quota history, cycle length, technical-buyer fluency) are extracted, scored, and auditable—without hallucinated claims.
• Partner with search leads to productize one “search OS” capability per quarter (territory graph, relationship density, or forecast-of-close) that recruiters actually use on live mandates.
• Establish data contracts, PII handling, and model-risk reviews so every candidate-facing and client-facing output is defensible in a relationship-driven staffing business.

Skills

Production NLP/IR systems

LLM applications with retrieval

PII handling

Domain rules to features

Principal IC delivery

Clear written communication

Tools

Vector stores

LLM platforms

Job description

Recruits Lab exists to place the right life sciences and technology talent—quietly, precisely, and faster than the market. We win searches because we map passive talent, verify named-account relationships, and deliver calibrated shortlists, not résumé piles. This role builds the AI that makes that discipline scale: systems that surface the unlisted expert, score fit against real hiring constraints, and give our partners a decision-ready slate in weeks, not quarters.

Performance Objectives

• Ship, within 90 days, a production retrieval and ranking pipeline that improves passive-candidate shortlist quality (precision at 10) by at least 25% versus current sourcing on two live life sciences searches.
• Design and own an evaluation harness that measures match quality against hiring-manager calibration notes, quota/territory constraints, and named-account verification—not keyword overlap—and report lift monthly.
• Reduce time-to-calibrated-shortlist on repeat search types by 30% within six months by automating mapping of competitor talent pools and adjacent buyer relationships.
• Put LLM-assisted screening into the recruiter workflow so structured signals (quota history, cycle length, technical-buyer fluency) are extracted, scored, and auditable—without hallucinated claims.
• Partner with search leads to productize one “search OS” capability per quarter (territory graph, relationship density, or forecast-of-close) that recruiters actually use on live mandates.
• Establish data contracts, PII handling, and model-risk reviews so every candidate-facing and client-facing output is defensible in a relationship-driven staffing business.

Environment & Resources

You report to company leadership in a 11–50 person, early-stage, for-profit team based in Fort Lauderdale, working remotely-first with recruiters who live in the search. You will have latitude to choose models, vector stores, and orchestration; production traffic is internal and client-confidential, not consumer-scale. Success is measured by placement speed and shortlist calibration, not vanity model metrics. You sit beside the people who close Area Sales Managers in Tri-State, Midwest, Southeast, and Northeast—so your systems must survive real geographic, technical-buyer, and passive-talent constraints.

Essential Qualifications

• Shipped production NLP/IR or ranking systems (search, matching, or recommendation) where quality was measured against human expert judgment, not only offline benchmarks.
• Hands‑on ownership of LLM applications with retrieval, structured extraction, evaluation, and guardrails for high‑stakes, PII‑sensitive workflows.
• Ability to translate messy domain rules (territory, quota, multi‑stakeholder cycles) into features, labels, and product behavior recruiters will trust.
• Proven delivery as a principal IC: architecture, code, evals, and stakeholder alignment without a large platform org underneath you.
• Clear written communication that matches how we work: specific, outcome‑based, no jargon theater.

Role Selling Points

• Your models sit on the critical path of real placements in biotech, pharma, digital pathology, and technical sales—not a sandbox.
• Tiny team, high agency: you define the AI stack for a firm whose edge is already data‑driven, candidate‑centric search.
• Domain depth most AI roles never see: passive maps, named‑account diligence, and hybrid technical‑commercial profiles.
• Work that compounds—every search you instrument makes the next calibrated shortlist sharper.