Headline

Agentic AI Engineer | Senior Data Engineer | Capgemini | Python | Pyspark | SQL | Microservices | LLM | RAG | Langchain | LangGraph | Agentic AI | MCP | A2A

About

Agentic AI Engineer and Data Engineering at Capgemini, specializing in production-grade Generative AI systems. I design and implement RAG and non-RAG LLM pipelines, LangGraph-based multi-turn agent workflows, MCP-integrated tool orchestration, and knowledge-driven agents powered by vector databases and Neo4j knowledge graphs for hybrid (Vector + Graph) retrieval systems. My experience includes building LLM evaluation and validation frameworks covering relevance, correctness, consistency, hallucination detection, and edge-case analysis—improving response precision and reducing downstream errors by ~25% while enhancing system reliability to 96%+ through monitoring, retry logic, and exception handling. I apply system-level design principles to agentic workflows, focusing on memory management, prompt versioning, retrieval optimization, turn tracking, and response evaluation. Beyond AI systems, I architect event-based and schedule-based orchestration workflows using AWS Step Functions and cloud-native services, design secure microservices with JWT authentication, RBAC authorization, and session management, and develop end-to-end Bronze–Silver–Gold lakehouse pipelines using Python, PySpark, Snowflake, AWS, and Azure. I also build OCR-driven unstructured data processing pipelines integrated into intelligent retrieval systems. Strong in Python, distributed data engineering, cloud platforms, and LLM foundations, I focus on building reliable, scalable, and measurable AI-data systems that deliver real-world business impact in Agile, cross-functional environments.

Location

Mumbai

AI stack

aistudio

langfuse

sora

v0

Work experience

Senior Agentic AI Engineer at Capgemini, Apr 2023 - Present · 3 yrs 2 mos, Mumbai, Project: IDEA SDLC (Agentic AI-IDEA) • Architected and deployed a multi-agent AI platform using AWS Bedrock (LLM) and LangGraph, enabling end-to-end SDLC automation through intelligent, multi-turn agent orchestration. • Designed specialized agents for automated code generation (multi-language support), BRD documentation, HLD generation, data mapping design, and synthetic data creation, accelerating development lifecycle efficiency by ~45%. • Implemented self-healing code agents with automated debugging, optimization, and standardization workflows, reducing manual rework by ~35% and improving code quality compliance. • Built hybrid RAG and non-RAG pipelines integrating vector embeddings and Neo4j knowledge graphs (Graph + Vector Retrieval), improving contextual accuracy by ~30%. • Developed memory-aware agent workflows with session persistence, conversation tracking, and long-term knowledge retention to enhance multi-turn reasoning consistency. • Engineered LLM evaluation frameworks covering relevance, correctness, hallucination detection, and edge-case validation—improving response precision by ~25% and achieving system reliability of 96%+. • Built embedding pipelines and semantic search layers using vector databases and Neo4j knowledge bases to support knowledge-driven chatbot interactions. • Integrated monitoring, retry logic, and automated guardrail enforcement to ensure safe, consistent, and production-grade LLM outputs. • Delivered an intelligent chatbot platform combining LLM reasoning, knowledge graph traversal, RAG retrieval, and agent collaboration to enable scalable AI-driven SDLC transformation.

Data Analyst at Capgemini, Apr 2023 - Present · 3 yrs 2 mos, Mumbai, Maharashtra, India, Project: Sustainability Data Hub(Azure-SDH) • Developed parameterized ADF pipelines for batch ingestion of data files from ADLS, utilizing a • • microservice-based ingestion framework. • Built ADF pipelines for stream ingestion using Kafka Confluent, enabling real-time Datasphere data migration from APIs to ADLS. • Implemented ADF pipelines for API-based ingestion of SAP Datasphere data into ADLS. • Designed and implemented ADF pipelines for data migration, coupled with Azure Databricks notebooks to perform transformations on ingested data. • Performed layered data processing (Bronze → Silver → Gold), incorporating cleansing, reconciliation, and profiling operations across each layer. • Stored all job-related metadata into MongoDB, capturing job registration and execution details. • Used Mobaxterm to monitor backend services, check pod statuses, and analyze logs.. • Stored Silver layer transformed data into PostgreSQL tables, and performed further transformation at the Gold layer, saving the final outputs into Gold tables in PostgreSQL. • Developed end-to-end microservices using Python and Flask to support metadata extraction from diverse sources and transformed it into visual formats for data analysis.

Data Analyst at Capgemini, Apr 2023 - Present · 3 yrs 2 mos, Mumbai, Maharashtra, India, Project: Sustainability Data Hud (AWS-SDH) • Developed microservices configuring SDH with source/target/intermediate file store connection details. • Developed Teradata-Redshift Stream microservices in SDH for schema conversion and object creation on AWS Redshift using regex. • Developed Teradata-Redshift microservice using Glue PySpark Script for diverse data loading (Full, Incremental, Adhoc) from source to staging and from staging to destination. • Developed Amazon Glue PySpark Script for diverse source data batch ingestion to S3 Bucket, enabling Full and Incremental Load operations. • Developed parameterized PySpark notebook templates on EMR for data transformation on ingested data. • Utilized SonarQube to analyze and mitigate code issues, achieving an 80% reduction in bugs and noise through code refactoring. • Managed code versioning with bit bucket and deployment to staging and production servers. • Utilize PyTesting, the Python unit test framework, for all Python applications.

Internship Trainee at Metalman Auto Limited, Jun 2021 - Sep 2021 · 4 mos, Aurangabad, Maharashtra, India, • Understand Requirements: Gather all necessary specifications and requirements for the sheet metal designs. • Draft Initial Designs: Use AutoCAD to draft the initial designs, ensuring they align with the sheet metal die specifications. • Detailing: Add detailed dimensions, tolerances, and annotations to the designs. • Review and Revise: Review the designs for accuracy and make necessary revisions based on feedback. • Import 2D Designs: Import the 2D designs created in AutoCAD into SolidWorks. • Create 3D Models: Use SolidWorks to convert the 2D designs into 3D models, ensuring all features and specifications are accurately represented. • Simulation and Analysis: Perform simulations and analyses to check the feasibility and performance of the 3D models. • Finalize Designs: Make any necessary adjustments and finalize the 3D designs for production.

Education

Bachelor of Engineering - BE, Mechanical Engineering, Sanjivani College of Engineering kopargaon

Bachelor of Engineering - BE, Mechanical Engineering, Savitribai Phule Pune University