Industrialize our retrieval and reasoning models — from embedding pipelines to evals to safe rollouts.
About the role
You sit between AI Engineering and Platform. You make sure models, embeddings, and pipelines ship safely, are evaluated continuously, and can be rolled back without drama.
What you'll do
- Own the embedding pipeline: ingestion, chunking, multi-embedding, reindexing
- Build eval harnesses and quality dashboards for retrieval and answer quality
- Operate Qdrant collections, index versioning, and zero-downtime reindexing
- Automate model/version rollouts with shadow traffic and canary patterns
- Track drift, latency, cost, and quality per domain (legal, financial, real estate)
What we look for
- 5+ years in MLOps / ML Platform / Applied ML Infrastructure
- Hands-on with vector databases (Qdrant preferred), embedding pipelines, and reranking
- Strong Python; solid with batch + streaming job orchestration
- Experience designing offline and online evals for retrieval/LLM systems
- Comfortable owning data + model lifecycle end-to-end
Nice to have
- Experience with Cohere / OpenAI / Gemini production usage at scale
- Background with feature stores, experiment tracking (MLflow, W&B)
- Prior work on RAG quality monitoring or hallucination detection
Stack & tools
Python
Supabase
Qdrant
MLflow
Airflow
Docker
OpenTelemetry
How to apply
Send a short note explaining why you're a fit, plus 1–3 concrete artifacts (code, writing, deals closed, products shipped). We read every thoughtful application.
