I'm Bhupalam Vishnu Varshith — a 2026 B.Tech AI & ML graduate building production-ready ML, Computer Vision, NLP, RAG and Agentic AI systems end-to-end.Open for freelance, project collaborations & fresher roles.
I'm a 2026 B.Tech graduate in Artificial Intelligence & Machine Learning from AITS Rajampet (CGPA 8.56), specialized in ML, Deep Learning, Computer Vision, NLP, RAG, and Agentic AI — with full-stack chops to ship them end-to-end.
I love turning research into reliable products: training CNN / RNN / GAN models, building hybrid retrieval systems (FAISS + BM25 + re-ranking), fine-tuning YOLO for real-time vision, and wiring it all into FastAPI, Flask, Spring Boot and Streamlit interfaces.
Open for freelance work, project collaborations, and fresher / new-grad AI & full-stack roles — feel free to reach out for anything in that space.
The technologies I use to ship reliable, performant, and scalable software.
A snapshot of the products, tools, and experiments I've built recently.
Final-year clinical ML pipeline on 12 patient-vital features. CatBoost (91.6%) and LightGBM (92.4%) with 8–12 ms inference. SHAP explainability in a Streamlit dashboard so physicians can trace risk scores to individual biomarkers.
Hybrid retrieval pipeline (FAISS dense + BM25 sparse + neural re-ranking) cutting hallucination ~30% vs naive RAG on 50-page benchmarks. Local LLM via Ollama with sub-2s end-to-end response on 8 GB consumer hardware — fully offline.
Fine-tuned YOLO on a 2,000-image custom dataset. Achieved 0.7823 mAP at 30 FPS real-time inference, deployed as a live Streamlit app for road-maintenance teams.
Fused text sentiment (RoBERTa + VADER) with a facial-expression CNN into one Flask REST API — real-time emotion scores from text, images and webcam at 84% cross-modal accuracy on a 1,000-sample test set.
Full-stack web app for university operations — role-based auth, admin panel, faculty, student, course and department management with a responsive UI.
Offline AI summarization platform that turns long blogs and PDFs into concise summaries using local LLMs (Ollama) and extractive techniques.
A short timeline of the path that shaped how I build today.
Live snapshot of my open-source activity, contributions, and ranking.
Open to freelance work, project collaborations, and fresher AI / full-stack roles — let's talk.