Margi Chaudhary

I’m an AI Developer intern building real-time computer vision systems that perform under real-world constraints. At iQud Tek, I developed a multi-stage AI pipeline for automated cleanroom smoke validation using computer vision and motion analysis on real-time video streams, reaching 90–92% accuracy and reducing manual inspection effort by ~60%.

I also design end-to-end edge-ready solutions and production-friendly services. From my work at IIT Roorkee, I built 3+ real-time tamper detection algorithms on Raspberry Pi 5 with sub-200ms inference latency, created data collection/annotation pipelines for 400+ images, and fine-tuned YOLOv8 for robust performance across varied lighting—while using GitHub for reproducible, metric-logged experiment workflows. Beyond internships, my projects include a YOLOv8-based assistive system with 92.4% detection accuracy across 20+ object classes and an LLM-powered museum ticketing chatbot (LLaMA 3.1 + OpenAI API) with 100%+ NL queries and 80%+ accuracy supported by an SQL data layer. I’m driven by challenging engineering environments where I can scale real-world vision models.