Data scientist - ai evaluation and responsible ai

Apply now

Location

onsite, Bucharest, B, Romania

Commitment

Full Time

Level

Senior (5+ years)

Required skills

Data ScienceMachine LearningPythonSQLResponsible AIModel ValidationRAGLLMMLOpsLLMOpsStatisticsExperimental DesignAdversarial TestingPII ValidationModel Risk ManagementTechnical Communication

Job Description

Who We Are

We operate beyond organizational silos, combining deep expertise in data engineering and AI-driven solutions to design and deliver scalable, user-centric applications and modern data architectures. We support a wide range of industries, including Automotive, Banking, Insurance, Telecommunications, E-commerce, and Public Services.

What You’ll Be Doing

  • Define evaluation strategies, golden datasets, metrics, thresholds and release gates for ML, RAG, agent, document and vision solutions.
  • Build automated evaluation harnesses, regression tests, red-team and prompt-injection tests, and human-evaluation protocols.
  • Define responsible-AI controls covering explainability, bias, PII, human oversight, traceability and model risk.
  • Integrate evaluation into CI/CD and MLOps or LLMOps pipelines.
  • Specify monitoring, alerts, model cards and audit evidence, and investigate production quality issues.
  • Work hands-on with use-case teams from baseline creation to release decisions and report quality against business measures.
  • Challenge weak evidence or unsafe release decisions constructively.
  • Deliver reusable, tested Python code, evaluation assets and governance templates.

What you'll bring along

  • At least 10 years of professional experience preferred
  • 6+ years of experience in data science, ML evaluation, model validation, experimentation or responsible AI.
  • 4+ years of strong Python and SQL experience with model, log and monitoring data.
  • Production experience evaluating and monitoring AI or ML systems.
  • Strong foundation in statistics, experimental design, error analysis and calibration.
  • Experience with RAG, LLM or agent evaluation and safety testing.
  • Ability to challenge weak evidence or unsafe release decisions constructively.
  • Strong technical communication, stakeholder influence, ownership, mentoring and judgment under ambiguity.
  • Experience with adversarial testing, PII validation, Bedrock Guardrails, CloudWatch, New Relic, Langfuse or safety-sensitive applications is preferred.

Excellent command of English, both written and spoken

Ready to join the team?

Apply now