Work Experience

Data Scientist at Rinventix Technologies

Insurance Document Digitization Platform

  • Built an automated digitization pipeline to process 10,000+ commercial insurance policy documents, extracting structured data from unstructured formats including PDFs, spreadsheets, and email attachments.
  • Developed a multi-modal extraction engine using LLMs, VLMs, and OCR to interpret unstructured data, complex tabular layouts, and diverse document types.
  • Developed an Evals Agent that systematically identified and corrected extraction errors, increasing system accuracy from 80% to 95% as validated by QA teams.

Real-Time Voice AI & Telephony Platform

  • Built an end-to-end voice AI pipeline using LiveKit (WebRTC) and Twilio to enable low-latency, bidirectional conversations between AI agents and users over web and phone (SIP).
  • Developed the core audio processing flow, integrating Speech-to-Text (STT), Text-to-Speech (TTS), and Voice Activity Detection (VAD) to handle natural speech interruptions and pauses.
  • Automated complex task execution by building AI Agents capable of triggering external API workflows and managing state during live voice interactions.

Advanced RAG & Information Retrieval System

  • Built a RAG pipeline using Hybrid Search (Vector + Keyword) and Reranking to optimize data retrieval from massive datasets of PDFs, Docx and web sources.
  • Implemented advanced retrieval techniques such as Query Expansion and HyDE (Hypothetical Document Embeddings) to significantly improve the relevance of context provided to LLMs.
  • Created a multi-source data ingestion engine capable of parsing and structuring information from diverse formats (Books, Websites, and structured JSONs) for semantic search.

Data Scientist at Aidetic Software

Content Discovery and Recommendation System

  • Built a data scraper using Selenium and Scrapy, and used OCR, Transformer models, and FAISS to create a fast vector search system for discovering and recommending relevant content.

Intelligent Video Analytics System

  • Built a real-time video analysis system on NVIDIA Jetson, setting up the full flow from camera feed to data insights for edge computing.
  • Developed age and gender detection models using CNNs and OpenCV, handling everything from face detection to image processing for demographic analysis.

Projects

MFReturns.com

Full-Stack Financial Tool

  • Built and launched a mutual fund analytics platform (FastAPI, SQL, Next.js) computing SIP, CAGR, and XIRR returns across 3,000+ schemes, backed by a weekly automated NAV data pipeline.
  • Deployed and maintained independently on a self-managed VPS with CI/CD automation.