Hi, I'm Himanshu Kumar Bhagat
AI/ML Engineer | LLM Builder | Skilled in Python, Deep Learning & Data Analysis
🚀 I build real-world, human-centric AI systems using deep learning, LLMs, and creative pipelines.
AI assistant that reads medical reports from PDF, image, or text, and explains them in simple, layman-friendly language. Supports multilingual voice output, follow-up questions, and real-time risk detection. Designed to serve low-resource rural populations.
End-to-end system that fetches real-time news, generates human-like scripts using Mistral, converts them into a realistic voice, and produces anchor-led avatar videos with B-roll visuals — all automatically.
Developed a transformer-based model (GPT-2) from scratch and fine-tuned it on a custom poetry dataset. Designed the tokenizer, training flow, and model behavior to generate rich, stylistically coherent poetry.
Built a hybrid CNN model (ResNet50v2 + custom layers) for rice disease detection from leaf images. Trained on 15,000+ samples and achieved 99.5%+ accuracy. Published in the IJSREM research journal.