
Sheetal J
AIML Engineer
Habilidades

Conheça meus serviços

Experiência profissional
AI Engineer
Nagarro SE
Mar 2026 - Present • 8 mos
Built production RAG applications with LangChain, LangGraph, and Azure AI Search, using hybrid retrieval (BM25 + vector + semantic ranker) with RRF fusion, metadata filtering, and security trimming. Designed agentic workflows for query routing, decomposition, and expansion (HyDE, multi-query), with self-corrective loops (retrieval grading, groundedness checks, bounded retries) and cross-encoder reranking for precision. Developed ingestion pipelines with Azure Document Intelligence, parent-child and semantic chunking, and versioned indexes, delivering source-grounded answers with citations and abstention on weak evidence. Automated evaluation with DeepEval (faithfulness, contextual precision/recall, G-Eval) as CI/CD regression gates, and cut latency and cost via semantic caching, model routing, and tracing with LangSmith/Langfuse.
Data Scientist
T-Systems International
May 2021 - Mar 2026 • 4 yrs 10 mos
- Built and deployed AI-driven solutions integrating ML, NLP, and Generative AI, supporting predictive analytics, decision automation, and enterprise use cases. - Designed and implemented RAG-based retrieval systems using LangChain, FAISS vector stores, embeddings, and cross-encoder reranking to enable accurate, low-latency LLM responses. - Worked with LLMs and transformer-based architectures to build context-aware applications using structured prompting, grounding, and retrieval pipelines. - Developed multi-agent workflows (MCP-style orchestration) enabling tool-based reasoning, retrieval coordination, and modular AI pipelines. - Implemented end-to-end ML pipelines including data preprocessing, feature engineering, model training, evaluation, and optimization for large-scale datasets (7M+ rows). - Deployed enterprise-grade AI systems to production on AWS, leveraging Dockerized services, REST APIs, CI/CD, and automated retraining workflows. - Applied MLOps best practices using MLflow for experiment tracking, model registry, versioning, and monitoring. - Delivered measurable business impact including $500K annual cost savings, 5% revenue retention improvement, and 12% reduction in storage costs. - Collaborated with cross-functional teams in ambiguous, fast-moving environments to translate business requirements into scalable AI solutions.