Santu Hazra

Technical Lead - Artificial Intelligence

JPMorganChase | June 2025 - Present

• InsightsAI – Agentic AI & RAG: Enabled 1,000+ users to search and generate insights across 2,000+ marketing reports by building an Agentic AI/RAG platform using LLMs, AWS OpenSearch, and ECS.
• Enterprise Knowledge Pipeline: Helped replace a ~$50M external research solution by implementing a RAG pipeline with automated daily ingestion from 5 vendors using AWS Lambda, EventBridge, OpenSearch, and ECS.
• MetaCortex – NL-to-SQL: Enabled natural-language analytics across 100+ enterprise tables by building a LangGraph-based NL-to-SQL agent with Graph DB schema discovery and automated insight generation.
• Production AI Analytics: Productionized MetaCortex on AWS, enabling 50+ users to conversationally query complex marketing data and generate actionable insights.
• MAP AI Automation: Leading a multi-agent AI platform targeting an 80–90% reduction in campaign cycle time by automating end-to-end marketing workflows.
• Multi-Agent Architecture: Designed an orchestrator and specialized agent architecture for 200+ marketers covering requirements, audiences, creatives, journeys, and campaign performance.

Senior Software Engineer - AI/ML

Machani Robotics | July 2024 - May 2025

• Multilingual Speech Recognition: Developed a real-time speech-to-text pipeline using GStreamer and Whisper CPP, supporting 5 languages for multilingual human-robot interactions.
• Whisper Fine-Tuning: Fine-tuned Whisper Base using 40+ hours of Common Voice and LibriSpeech datasets, reducing Word Error Rate by 15%.
• LLM Fine-Tuning: Fine-tuned Phi-3 (3B) using 2,000+ domain-specific Q&A pairs generated with GPT-4 to improve contextual conversations.
• Retrieval-Augmented Generation: Designed a RAG framework combining LLMs with face and voice context for personalized real-time conversations.

SDE-I - AI Engineer

Machani Robotics | February 2021 - June 2024

• Face Recognition: Built an ArcFace-based recognition pipeline achieving 92% accuracy, optimized with DeepStream and ONNX for NVIDIA AGX Jetson.
• Vector Search: Developed FAISS and Milvus-based vector search for real-time face and speaker identity verification.
• Text-to-Speech: Integrated TTS systems supporting emotion and voice cloning, improving voice clarity and contextual relevance by 20%.
• Gesture Generation: Developed an LLM-driven gesture generation system, improving humanoid body-language and expression relevance by 30%.
• Edge AI: Optimized ONNX models and GStreamer pipelines, reducing inference latency by 20% on NVIDIA Jetson hardware.

Deep Learning and AI Instructor (Part Time)

AnalytiixLab | April 2022 - Present

• Trained 150+ students across 8 batches in Deep Learning and Artificial Intelligence.
• Conducted corporate AI training programs for Tredence Inc. and Samsung.

Associate Data Scientist

Cognizant Technology Solutions | April 2015 - January 2021

• Predictive Analytics: Developed churn prediction and lead-scoring models, improving customer retention and acquisition efficiency by 15–20%.
• NLP: Performed sentiment analysis across 10,000+ customer reviews to generate actionable product and marketing insights.
• Computer Vision: Built driver-behavior classification models using dashcam imagery, reducing incident detection time by 25%.

B.Tech In ECE

West Bengal University of technology | August 2010 - July 2014

Higher Secondary

West Bengal Board of Higher Secondary Education | August 2007 - July 2009

Secondary

West Bengal Board of Secondary Education | July 2007

Certifications

• Wiley Data Scientist. Credential ID: CZN-CDS-BAN-210819004
• Extensive Vision AI program. from The School of AI
• Extensive Python AI program. from The School of AI
• Extensive NLP program. from The School of AI