Ronit Virwani
AI Engineer • CS @ Binghamton • Building Retrieval‑Augmented AI
- Role
- Ai Engineer at Akos
- Location
- Pittsburgh, PA, US
- LinkedIn followers
- 500 followers
About Ronit Virwani
Hi, I’m —a software engineer who likes turning machine‑learning…
Experience
Ai Engineer
May 2025 — Present · Phoenix, AZ, US
Built an automated petition processing system on Palantir Foundry that takes scanned PDFs and extracts individual signatures using DETR/YOLO models for table detection, Google Vision OCR for text extraction, and Gemini LLM for parsing signature data into structured records- Developed GPU-accelerated row cropping pipeline using PaddleOCR to detect individual signature lines within petition tables and automatically crop them into separate images for verification workflows- Created voter verification system that matches petition signatures against databases using both semantic similarity (Azure OpenAI embeddings) and exact string matching, with configurable confidence thresholds for different accuracy requirements- Implemented state-specific processing logic to handle different petition formats with adaptive algorithms that adjust detection parameters and cropping strategies based on document layout variations- Built a study guide generator using FastAPI and Next.js that processes 900+ ophthalmology textbooks and generates personalized study guides for medical residents- Applied Retrieval-Augmented Generation (RAG) with ChromaDB to search millions of medical document chunks, featuring priority collection system for customized searches based on specialty focus- Developed multi-LLM support (OpenAI GPT-4, Google Gemini, Llama) with Chain of Thought pipeline that breaks complex medical queries into focused sub-questions and synthesizes study guides- Built real-time WebSocket streaming with live progress tracking and chunked data transmission for handling large study guide responses- Created PDF processing pipeline that extracts text, images, and diagrams from medical textbooks, applies OCR for captions, and converts everything into searchable vector embeddings- Added medical abbreviation expansion system and section-level regeneration feature that allows selective improvement of study guide sections using custom prompts while preserving document structure
Education
Binghamton University
Master of Science - MS
2024 — 2026
MIT ADT University
Bachelor of Technology - BTech
2020 — 2024
Find verified contacts for anyone on LinkedIn
Unifers gives sales teams verified emails and direct dials, enriched profiles, and outreach that lands in the inbox.
Free plan included · No credit card required
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.