Lead Data scientist

Job description

**Your role:** The **Lead Data Scientist** **architects, builds, and runs** **production-grade** **Machine Learning** and **Generative AI** systemsowning the **full lifecycle** from **model development** to **scalable cloud deployment** and ongoing **performance monitoring**. In addition, the role partners with **commercial stakeholders** translate market/customer data into **decision-ready insights** and **AI-enabled analytics** solutions that drive **measurable outcomes** Operating with a **builder and translator mindset**, the individual rapidly develops **MVP analytics solutions**, leverages **AI** to accelerate **insight generation**, and ensures strong product engineering fundamentals, **data quality**, and **governance**. The role plays a critical part in establishing a **single source of truth** for **performance management** across markets and channels while elevating analytics maturity from descriptive reporting to **predictive** and **insight-led decision making**. **Key Responsibilities** **1) ML & Deep Learning Model Development** - **Design**, **train**, and **optimize** **ML models** for prediction, classification, ranking, **time-series forecasting**, **anomaly detection**, **NLP**, and **recommendation** use cases. - Build robust **experimentation workflows** (**train/validation strategy**, ablations, **error analysis**) and improve **model quality** through iterative tuning. - Ensure **reproducibility** and **maintainability** through clean code practices, **versioning**, and **automated testing**. **2) GenAI Engineering (LLMs, RAG / MCP / fine-tuning, Agents)** - Build enterprise-grade **LLM applications** using **RAG** (retrieval-augmented generation), **MCP**, and **fine-tuning** approaches: **chunking** strategies, **embedding generation**, **hybrid retrieval**, **reranking**, **prompt templates**, and **citation/attribution** patterns. - Develop **LLM applications** with **tool use**/**function calling** patterns and **agentic workflows** where appropriate. - Implement **systematic evaluation**: curated **eval sets**, **prompt regression tests**, **hallucination checks**, **retrieval quality metrics**, and automated **quality gates**. **3) ML & LLM Operations: Productionization, Deployment & Monitoring** - Deploy and operate **real-time** and **batch inference** solutions on **Azure** using **managed endpoints** and/or **containerized serving**. - Build **CI/CD** for ML systems: **automated packaging**, **container builds**, **model validation tests**, **staged rollouts**, and **rollback strategies**. - Establish **lifecycle management**: **model registry**/versioning, **lineage**, **promotion workflows**, and **release governance**. - Implement **observability**: **latency**, **throughput**, **cost**, **drift** signals, **data quality checks**, **alerts**, and **performance degradation** monitoring. **4) Pipeline Orchestration & Automation (Train Deploy)** - Build **standardized ML pipelines** for **training**, **evaluation**, and **deployment** using **orchestration tools** (cloud-native pipelines and/or platform tools). - Automate **dataset/version management**, **feature generation**, **scheduled retraining** triggers, and **approval workflows**. - Define **repeatable patterns** for **scalable experimentation** and **reliable production delivery**. **5) Analytics Products, Dashboards & Data Governance** - Own key **analytics outputs** as **products** (**dashboards**, **reusable datasets**, **internal tools**), continuously improving them based on usage patterns and **performance gaps**. - Build and automate **dashboards** and analytical components using scalable **SQL** logic, **Python** transformations, and **reusable modules**. - Act as owner for critical **commercial/syndicated datasets** (e.g., GfK, Circana, Nielsen or equivalent): **definitions**, **assumptions**, and **limitations**, ensuring **transparent logic** and **trust** in outputs. - Partner with **data engineering/IT** to ensure **data quality**, **harmonization**, and **governance** through strong **validation** and **reconciliation** practices. **6) Stakeholder Partnership & Decision Support (Lightweight, High Impact)** - Serve as **trusted analytics thought partner** to **senior stakeholders** (e.g., BU leadership, Sales, Marketing, Finance), shaping **problem statements** and aligning on **success metrics**. - Translate complex analytics into clear **recommendations** with a **decision-oriented storyline** (“**so-what / now-what**”), tailored for **leadership forums** and **reviews**. - Support **performance reviews**, **planning cycles**, and high-priority **ad-hoc requests** with **speed**, **rigor**, and **confidence**; proactively **challenge assumptions** with **fact-based insights**. **7) Responsible AI, Security, and Risk Controls (GenAI-ready)** - Implement **guardrails**: **prompt injection** defenses, **sensitive data** protections, **output validation**, and **secure tool execution** patterns. - Apply **responsible AI** practices: transparent **evaluation criteria**, **auditability**, and **risk controls** aligned to enterprise needs. **8) Technical Leadership (Lead-level Expectations)** - Set **engineering standards** for DS/ML codebases: **design docs**, **code review** practices, **testing discipline**, and **production readiness** checklists. - **Mentor** data scientists/ML engineers on **modeling**, **GenAI engineering**, and **MLOps** best practices. - Lead **architectural decisions** across modeling approaches, **retrieval stack**, **serving patterns**, and **evaluation strategy**. **Core Skills & Competencies** **Must-have (Technical)** - Strong **Python** (**production-quality** coding) and solid CS fundamentals; strong **SQL** for **data access** and **validation**. - Depth in **ML**: Traditional **ML** exposure and at least one **deep learning** framework (**PyTorch/TensorFlow**), with strong understanding