01Overview
Data Scientist (Senior Lead)
Experience: 15 to 20+ years
Function: AI & Machine Learning
Location: Cognizant India (Hybrid)
ROLE SUMMARY
Owns the ML and GenAI model strategy for the product. Leads RAG design, hybrid retrieval, model evaluation with LLM-as-a-Judge, and LLM fine-tuning programmes. Sets the standard for how agents retrieve, reason, and generate reliable outputs.
KEY RESPONSIBILITIES
- Define and own the ML/GenAI model strategy for all agentic use cases on the platform
- Lead RAG pipeline design, hybrid retrieval strategy, and vector/graph database selection
- Establish model evaluation frameworks: benchmarking, accuracy, LLM-as-a-Judge calibration
- Lead fine-tuning programmes: LoRA, QLoRA, PEFT, DPO, SFT, RFT (Reinforcement Fine-Tuning)
- Guide the ML Engineer on MLOps practices, model promotion, and serving optimisation
- Run structured A/B tests, model comparisons, and prompt strategy evaluations
- Produce model cards, evaluation reports, and executive-level performance insights
TECHNICAL SKILLS
- Languages & Libraries: Python; PyTorch, TensorFlow, scikit-learn, Hugging Face Transformers, LlamaIndex
- LLMs & Fine-Tuning: LoRA, QLoRA, PEFT, RFT, RLHF, DPO, SFT, instruction tuning
- RAG & Hybrid Retrieval: BM25 + dense vector, reranking (Cohere Rerank, ColBERT, cross-encoders)
- Vector & Graph Databases: Pinecone, pgvector, Weaviate, Qdrant; Neo4j, Memgraph
- Evaluation: Ragas, DeepEval, HELM, PromptFoo, TruLens; LLM-as-a-Judge design and calibration
- MLOps: MLflow, Weights & Biases, experiment tracking, model registry, drift monitoring
- Cloud AI: AWS SageMaker, Azure ML, GCP Vertex AI
- Statistics: experiment design, A/B testing, causal inference, significance testing
NICE TO HAVE
- Published research at NeurIPS, ICML, ACL, or equivalent
- Agentic AI framework experience (LangGraph, AutoGen)
- Responsible AI tooling: Fairlearn, AI Fairness 360, Giskard
- Knowledge graph and LLM knowledge pipeline design experience .
