Purushottam Bende

I’m a B.Tech student in Artificial Intelligence (Aug 2022 – May 2026) and an AI/ML intern focused on shipping practical, measurable ML solutions. I enjoy turning messy inputs—documents, images, and real-world constraints—into reliable outputs people can act on.

As an AIML Intern at Remote Maharashtra Remote Sensing Application Centre (Dec 2025 – May 2026), I built an end-to-end AI system fine-tuning Mistral-7B (7.3B params) with LoRA on Maharashtra land records, reaching 86% accuracy in Marathi with training loss reduced from 2.376 to 0.051. I also built an OCR pipeline using EasyOCR + OpenCV to process 173 government PDFs and extract 600K+ characters, using MD5 caching to make queries 2× faster.

My project work spans analytics and computer vision: I created an interactive Superstore Sales dashboard in Power BI with time-series analysis (2014–2017), and I built a Security Kit Detection System using YOLOv11 to detect safety equipment with automated violation alerts and analytics logging. I also developed a medical imaging pipeline for brain tumor classification and segmentation using CNN and U-Net, achieving >92% classification accuracy with high-precision masks.

I’m especially motivated by systems that close the loop between detection and action. I had a PPE compliance detection project accepted at an international conference (AI-Driven Engineering & Technology-2025), where I used a YOLOv11 person-first pipeline with Twilio-based automated alerts for rapid supervisory action.