#SeniorDataScientist & AI #ArchitectGenerativeAI

Agentic Systems | Cloud-Native ML | Enterprise AI Experience 6+ Years Notice Period: Immediate Joiner

Location: Remote / Consultant Interested candidates can share cv to [email protected] KEY RESPONSIBILITIES • Architect and implement Agentic AI systems using multi-agent orchestration patterns (ReAct, Plan-and-Execute, Reflection, Tool-Use).

- Design and optimize Retrieval-Augmented Generation (RAG) pipelines including hybrid search, re-ranking, and chunking strategies.
- Build production-ready LLM-powered applications leveraging Claude, GPT, and open-source transformer models.
- Develop and maintain evaluation frameworks using Ragas, LLM-as-a-Judge, and custom benchmarks to measure RAG & agent quality.
- Implement end-to-end tracing and observability solutions using Langfuse and LangSmith for prompt lineage, latency, and cost monitoring.
- Lead Model Context Protocol (MCP) and Agent-to-Agent (A2A) integrations for composable, interoperable AI workflows.
- Deploy and manage AI workloads on AWS (Bedrock, Agent Core, OpenSearch, SageMaker, Lambda, EC2) with robust DevOps pipelines.
- Apply advanced prompt engineering techniques including ReAct, Chain-of-Thought, Few-Shot, and Reflexion prompting.
- Drive ML lifecycle management: feature engineering, model training,



experiment tracking, A/B testing, and production monitoring.
- Collaborate with Pharma, Banking, and Supply Chain stakeholders to translate business problems into scalable AI architectures.
- Champion responsible AI practices: fairness, explainability, data privacy, and regulatory compliance (HIPAA, GDPR).
- Containerize and orchestrate services using Docker, Kubernetes, and implement CI/CD pipelines for model deployment. MUST-HAVE TECHNICAL SKILLS Agentic AI & Generative AI • RAG: Advanced RAG patterns — naive, modular, graph RAG, multi-hop, hybrid retrieval, re-ranking (Cohere, cross-encoders).
- Agentic AI Patterns: ReAct, Plan-and-Execute, Reflection, Critique, Tool-Use, Router, Orchestrator-Subagent patterns.
- Transformers: In-depth understanding of attention mechanisms, BERT, GPT architectures, fine-tuning (LoRA, QLoRA, PEFT).
- Evaluation: Ragas framework, LLM-as-a-Judge, G-Eval, BLEU/ROUGE, faithfulness, answer relevancy, context precision metrics.
- Tracing: Distributed tracing of LLM calls, prompt/response logging, token usage, hallucination detection pipelines.
- Agentic Frameworks: LangChain, LangGraph, CrewAI, AutoGen, Llama-Index, Semantic Kernel — production-grade usage.

📌 Senior Data Scientist (Maharashtra)
🏢 Follex Technology Solution
📍 Maharashtra