AI Data Scientist: Evaluation & Retrieval for Orders

Volta Software; Lombardia; Hybrid; EUR 70,000 - 110,000; Full time

Job summary

Volta Software is seeking an Applied Data Scientist to measure and improve the accuracy of its AI system across real purchase orders. You will build evaluation harnesses, perform error analysis, and enhance retrieval and extraction from complex documents.

You will work with Python, Pandas/Polars, and LLM APIs to turn varied data sources (emails, PDFs, spreadsheets) into clean, structured order lines, with rigorous per-field metrics and CI-integrated tests. Hybrid Milan/Paris location preferred.

Qualifications

• 3+ years shipping ML or NLP to production.
• Strong Python, pandas or polars and testing habits.
• Hands-on with LLM APIs (Gemini, OpenAI, Anthropic or similar).
• Evaluation datasets and metrics (precision/recall, per-field accuracy, confidence calibration).
• Search and retrieval (Elasticsearch/OpenSearch, BM25, embeddings, reranking).
• Solid SQL.
• Fluent English; French or Italian.
• Nice to have: GCP (Cloud Run, Vertex AI); LLM tracing/evaluation tooling.

Responsibilities

• Own the accuracy number across customers, per field, per release.
• Build evaluation harnesses and CI-integrated tests for regression protection.
• Improve product and customer matching using Elasticsearch-based retrieval.
• Turn emails, PDFs, and spreadsheets into clean order lines with prompts and schemas.
• Perform rigorous error analysis to identify root causes and fixes and verify in production.
• Turn customer fixes into reusable templates and configurations by industry.

Skills

Python

Pandas

Polars

LLM APIs

Evaluation metrics

SQL

English fluent

French or Italian

Education

MSc in Computer Science/Statistics/Applied Math

Tools

Elasticsearch/OpenSearch

SQL databases

LangFuse (tracing)

Job description

Volta Software is seeking an Applied Data Scientist to measure and improve the accuracy of its AI system across real purchase orders. You will build evaluation harnesses, perform error analysis, and enhance retrieval and extraction from complex documents.

You will work with Python, Pandas/Polars, and LLM APIs to turn varied data sources (emails, PDFs, spreadsheets) into clean, structured order lines, with rigorous per-field metrics and CI-integrated tests. Hybrid Milan/Paris location preferred.