ML Engineer — Search, Ranking, and Eval

Role at a glance

Search quality is the product. Prove it in metrics.

Lumen Fintech (demo) powers internal research tools. You will build ranking and retrieval models, own evaluation harnesses, and work with product when “just tweak the model” is not a strategy.

What you will do

• Build reproducible training/eval loops: data snapshots, leakage checks, and canary release criteria tied to NDCG and business KPIs
• Partner with data eng on feature stores and with legal on PII: redact before it hits notebooks
• Ship model cards and dashboards PMs can read: what changed, what did not, and the counterfactual when we roll back

What we need

• 3+ years applied ML with ranking or retrieval; you have shipped, measured, and rolled back models
• You write code: training pipelines, not just notebooks, with tests around tricky transforms

How we will interview you

Paper discussion on an evaluation gap we hit in prod, a modeling exercise with our dataset (sanitized), and a stakeholder role-play on explaining a model limitation to a PM.

Note: this posting is demo data for the portal. Compensation band tests filters only. Apply via example.com addresses in this record.