Note: The job is a remote job and is open to candidates in USA. Careerminds provides career transition, coaching, and career progression solutions that support workforce growth and development. The Machine Learning Engineer will own machine learning products end to end, including discovery, design, development, deployment, experimentation, evaluation, and production outcomes across data resolution, retrieval, ranking, matching, LLMs, and agentic systems.
Responsibilities
- Canonical datasets for titles, companies, skills, and industries — the layer every application depends on. Content-addressed IDs, faceted taxonomies, alias graphs accumulated across tens of millions of rows
- Rules-based resolution pipelines with LLM escalation, where the accumulated alias graph is the durable asset and escalation volume should fall over time
- Nightly agent loops that adjudicate ambiguous entities, propose structural changes, and get gated by invariant checks and blast-radius limits before anything commits
- Job ingestion at scale: multi-source feeds, deduplication, freshness, and the indexing economics underneath
- Job matching v2: two-tower retrieval with cross-encoder reranking, trained on outcome labels rather than clicks. Hard-negative mining, propensity weighting, impression-time logging
- Mobility embeddings learned from observed career sequences — the similarity a text encoder can't recover, where Claims Adjuster and Underwriting Assistant are substitutable despite sharing no vocabulary
- Pivot feasibility: given where someone is, what moves are realistic, what's missing, and which intermediate roles actually worked for peers
- Fine-tuning where it earns its cost — against outcome labels, not for tasks a well-prompted frontier model already handles
- Agentic systems in production with human approval gates: agents that analyze, propose changes as reviewable artifacts, and execute only after a human signs off. We have this pattern running against tens of millions of customer touchpoints a year and want to push it much further
- Continuous skills inference from work artifacts rather than static documents — a problem several of our enterprise customers are currently solving for themselves, badly
- New product surfaces where the right answer genuinely requires an LLM, and the discipline to notice when it doesn't
- Evaluation infrastructure you'd defend in a design review: time-forward splits, calibration, offline-to-online agreement, and honest handling of feedback-loop degeneration and survivorship bias
- Building inside real constraints: GDPR, EU AI Act high-risk classification for employment AI, and client data commitments are design inputs here, not someone else's problem
Skills
- 5+ years shipping ML systems into production — and you can name the system, the metric before and after, and how you knew the model caused the change
- Depth in both classical ML and deep learning (PyTorch or TensorFlow) applied to live products, not notebooks and Kaggle sets
- Working fluency with LLMs in production — retrieval, evals, prompt and context engineering, and the judgment to recognize when an LLM is the wrong tool
- You already ship with agentic coding tools — Claude Code, Claude Design, or close equivalents — and can point to work you built with them
- Software engineering fundamentals strong enough to own your own deploys — Python, Git, cloud (we run AWS), containers, and the patience for genuinely messy, human-authored, self-reported data
- Entity resolution, record linkage, or taxonomy design at scale
- Ranking, recommendation, or two-tower retrieval systems
- Sequence models on longitudinal or event-stream data
- Embedding and vector retrieval systems in production
- Experiment design, causal inference, or off-policy evaluation
- Warehouse-native ML (dbt, Snowflake, or similar)
- Labor market, HR tech, or people-data domain experience
- Open-source contributions or publications
Benefits
- This is a 100% remote/work-from-home role.
Company Overview
Company H1B Sponsorship