Saksham Shinde

I’m currently a Data Scientist (Data Scientist - Manager) at HDFC, where I laid the technical foundation for production-grade RAG-powered AI engines deployed across major business units. By scaling client processing capacity from 20% to 85%, I earned an accelerated double-promotion directly to Manager.

I lead the design and implementation of core backend processing pipelines that automate complex data workflows and unblock downstream client-facing analytics models. I also architected an automated funnel analytics framework across 200+ transactional journeys, eliminating tracking dependencies and expanding per-engineer portfolio oversight by 300% (from 3 to 8–10 portfolios per person).

Before HDFC, I worked as a Junior Data Scientist at ODIO, where I fine-tuned Wav2Vec2-XLSR for Automatic Speech Recognition (ASR) across 8 regional languages. I achieved 89% character accuracy and 10% Word Error Rate (WER) and scaled training across distributed multi-GPU A100 workflows, while also building a real-time bi-directional customer-agent streaming framework using Flask, Flask-SocketIO, and FastAPI for low-latency audio processing and LLM orchestration.

My project work mirrors how I operate in production: end-to-end systems that are fast, reliable, and measurable. I’ve built a production-grade real-time conversational AI backend (Whisper, OpenAI API, AWS Polly, WebSockets, FastAPI) and an automated document ingestion & processing platform using CrewAI, LangChain, BERTopic, and LDA—improving file processing speed by 40% and raising semantic alignment and precision by 30%.