Machine Learning Engineer

hiair.ai; San Francisco (CA); Hybrid; USD 150,000 - 190,000; Full time

Benefits offered by this job

Equity

Job summary

hiair.ai is seeking a Machine Learning Engineer to own the matching engine and the ranking models behind every job we surface. This hybrid role sits at the heart of the recommender system, with responsibility for embeddings, the scorer, and maintaining scores across refresh sweeps.

You'll improve retrieval quality through offline evaluation and collaborate with apply agents to close the loop between recommendations and conversions.

Qualifications

• Solid ML/IR fundamentals: embeddings, retrieval, ranking, and how to evaluate them honestly.
• Strong Python and comfort with production data systems (Postgres, vector search).
• Experience taking a model or scoring system from notebook to production.

Responsibilities

• Own the embedding pipeline, the scorer, and the sweep that keeps match scores current.
• Design and run offline evaluations for retrieval and ranking; ship improvements against real metrics.
• Work with pgvector/Postgres at scale and keep the scoring path fast and cheap.
• Close the loop with conversion data so recommendations get better over time.

Skills

ML fundamentals

Python

Postgres

Vector search

Embeddings

Production data

Evaluation metrics

Tools

pgvector

Job description

Own the matching engine and the models that rank every job we show.

Hybrid

$150,000–$190,000 + equity

The matching engine is the heart of Hiair: it scores every job for every candidate on a 0–100 scale and decides what we surface. As our Machine Learning Engineer you own that engine — profile and job embeddings, the scorer, and the periodic sweep that keeps scores fresh.

You'll improve retrieval and ranking quality with real offline evaluation, not vibes, and partner with the apply agents to close the loop between what we recommend and what actually converts.

What you'll do

• Own the embedding pipeline, the scorer, and the sweep that keeps match scores current.
• Design and run offline evaluations for retrieval and ranking; ship improvements against real metrics.
• Work with pgvector/Postgres at scale and keep the scoring path fast and cheap.
• Close the loop with conversion data so recommendations get better over time.

What we’re looking for

• Solid ML/IR fundamentals: embeddings, retrieval, ranking, and how to evaluate them honestly.
• Strong Python and comfort with production data systems (Postgres, vector search).
• Experience taking a model or scoring system from notebook to production.
• Pragmatism about when a simpler model beats a fancier one.