At Shell India Markets, via Edunet Foundation, I trained a multi-class garbage classification model using TensorFlow and CNNs. It achieved 93% validation accuracy on more than 1,500 images, and data augmentation and hyperparameter tuning improved performance by 18%.
I built JobIQ, an AI career aggregation platform that brings together jobs, internships, hackathons, and competitions across India. Its matching system uses TF-IDF and cosine similarity to score candidates against postings.
For JobIQ, I also implemented an AI mock interviewer with 300 curated questions across 12 subjects, plus real-time speech-to-text and keyword-based scoring. I engineered scrapers for four platforms and deployed the application pipeline using Vercel, Render, and Supabase.
I also designed and deployed a Streamlit waste-classification app with a Python inference backend, image-upload interface, and real-time predictions. The 93%-accuracy classifier is hosted on Hugging Face Spaces.
