S.Deepan Chakravarti
AI Manager
Chennai,TN
Work Preference
Desired Job Title
AI ManagerForward Deployed EngineerWork Type
Full TimePart TimeConsultingLocation Preference
On-SiteRemoteHybridLocation: Chennai, TN, IN
Open to relocation: Yes
Summary
Results-driven AI Manager with extensive expertise in architecting and
scaling enterprise-grade intelligent document processing systems.
Proven track record of leading classification teams, migrating complex
ML/DL models to advanced Large Language Models (LLMs), and driving
modern engineering practices via AIDLC, agentic coding. Highly skilled in
designing robust RAG frameworks, fine-tuning LLMs, implementing
hybrid search techniques, and deploying cloud-native microservices on
AWS and Kubernetes to deliver scalable automation and high-precision
data extraction.
Overview
16
16
years of professional experience
Work History
Team Lead
Intellect Design Arena
Chennai, Tamil Nadu
01.2021 - Current
- Team leadership and delivery: Lead the document classification team, owning delivery, production support and stakeholder communication. Introduced AI-assisted engineering practices (AIDLC, Claude Code, Kiro, agentic coding) across the team.
- ML-to-LLM migration: Leading the migration of classification and extraction from traditional ML/DL models to LLM-based and agentic workflows, including LLM load balancing and hybrid approaches, for both native and scanned PDFs.
- Insurance document intelligence: Built end-to-end pipelines to classify and extract data from ACORD forms, loss runs, emails, broker specifications, supplemental applications, E-Mod, financial documents, quotes, binders, policies and broker-of-record letters. Added domain-specific classification for property underwriting (RMS, ATC, construction and occupancy codes).
- RAG and fine-tuning: Designed retrieval-augmented generation architectures with advanced chunking, hybrid sparse (TF-IDF, BM25) and dense (vector embedding) retrieval, and re-ranking; fine-tuned LLMs and built hybrid models for domain accuracy.
Prompt engineering: Applied structured prompt design and chain-of-thought reasoning to improve the quality and consistency of LLM outputs. - Document processing and model development: Built classification, extraction and checkbox-detection models with PyTorch, OCR (Tesseract, AWS Textract, Azure Document Intelligence), spaCy NER and YOLO. Evaluated models with AUC-ROC, confusion matrices, precision and recall, and automated production feedback pipelines for continuous improvement.
- Cloud deployment: Deployed microservices on AWS Lambda and containerised components on Kubernetes, using S3 for storage and SQS for messaging, to support scalable production use.
- Recognition: Received the company's Gem Award (March-2023) for integrating ACORD, loss run and broker specification processing into a deliverable product.
Senior Associate
Cognizant
Chennai, Tamil Nadu
07.2016 - 01.2021
- Big Data Migration: Led migration of Big Data to Microsoft Azure, enhancing data handling and storage efficiency.
Script Rewriting: Converted legacy scripts (Perl, Powershell, Oracle SPs) to Python, using advanced Python concepts for future-proofing. - Codebase Restructuring: Simplified and optimized the codebase, reducing redundancy and improving maintainability.
- Testing Frameworks: Implemented automated testing frameworks (Unit Test, Pytest, Tox), minimizing manual testing efforts.
- Azure Expertise: Developed data pipelines using Azure Data Factory, ADLS Gen2, Synapse, and Databricks for complex transformations.
- Python & Big Data: Used PySpark, RDDs, and Hive for data manipulation, and integrated Python for seamless operations across databases.
- SDLC & Agile: Followed Agile methodology, participated in Scrum calls, and managed deployment artifacts using Azure Pipelines and Boards.
Project: Metlife VADH (variable annuity dynamic hedging)
Software Developer
Virtusa Polaris
Chennai, Tamil Nadu
12.2010 - 07.2016
- IRaMS Development: Led the development of IRaMS, a web-based incident management system, overseeing the full SDLC from design to production support.
- Seamless Deployment: Ensured smooth deployments and incident management for BT's core platforms.
Business Analyst Role: Collaborated with clients to capture requirements and design low-level solutions for telecom incident resolution. - Knowledge Transfer: Facilitated knowledge transfer sessions for development teams to improve processes.
Legacy Migration: Led migration from legacy SQL systems, contributed to new feature development, and managed server migrations. - System Reliability: Improved system reliability by implementing housekeeping scripts and resolving critical issues.
Virtusa, BT (British-Telecom) TMW IRaMS (Incident Reporting and Management System)
Skills
Team leadership
Stakeholder management
Project lifecycle management
LLM classification & extraction
Agentic workflows
RAG & hybrid search
LLM fine-tuning
Prompt engineering
Deep learning
NLP
Computer Vision
OCR: Tesseract, AWS Textract
Model evaluation (AUC-ROC, precision, recall)
Core & Advanced Python
AWS: Lambda, S3, SQS
Docker, Kubernetes
Timeline
Team Lead - Intellect Design Arena
01.2021 - Current
Senior Associate - Cognizant
07.2016 - 01.2021
Software Developer - Virtusa Polaris
12.2010 - 07.2016