About this role
Proacure blends AI with domain expertise so procurement teams get trustworthy spend and sourcing insights. You will design, train, and ship models and data workflows that stay accurate in messy enterprise data—working with product, design, and customer teams from prototype to production.
You care about measurement: offline metrics, online evals, and human-in-the-loop feedback loops. You write code that other engineers can extend, and you are comfortable explaining trade-offs to non-ML stakeholders.
What you'll do
- Design and improve models and pipelines for classification, extraction, retrieval, and ranking over procurement documents and spend data.
- Build evaluation harnesses (golden sets, regression tests, LLM-as-judge where appropriate) and monitor drift and quality in production.
- Partner with full-stack engineers on APIs, batch jobs, and real-time features; optimize latency and cost for inference at scale.
- Collaborate with customer-facing teams on feedback, error analysis, and explainability so outputs are actionable for finance and procurement users.
- Document assumptions, datasets, and model cards; participate in design reviews and incident response when model behavior is in scope.
Skills and technologies
- Python
- PyTorch or TensorFlow
- SQL
- LLM prompting & RAG patterns
- Data pipelines (batch + streaming basics)
- REST / gRPC integration
- Git
- Cloud (AWS or GCP) fundamentals
Qualifications
- 3+ years shipping ML or applied AI in production (not only notebooks or Kaggle).
- Strong CS fundamentals: algorithms, probability, and systems thinking.
- Experience with NLP, tabular ML, or retrieval systems in a B2B or regulated context is a plus.
- Bachelor’s in CS, ML, or equivalent experience; advanced degree welcome but not required.