Darshan Shirsat

Machine Learning Engineer with an M.Tech in Artificial Intelligence & Data Science and hands-on experience building ML solutions across Industrial AI, Computer Vision, NLP, and MLOps. Experienced in developing end-to-end pipelines involving data preprocessing, feature engineering, leakage-aware evaluation, and deployment. Strong in Python, SQL, Scikit-learn, PyTorch, XGBoost, LightGBM, and practical machine learning workflows.

AI/ML Intern at CropGen (2025-09 – 2025-11)

• Developed a hierarchical crop disease classification system using EfficientNet-B0 for a large-scale agricultural image classification problem spanning 100+ crops and 6,000+ crop–disease combinations.
• Designed a two-stage computer vision pipeline combining crop identification with crop-specific disease classification for fine-grained prediction.
• Built a prototype covering 10 crop categories and 100 disease classes, achieving ~90% crop classification accuracy and 70–80% disease classification accuracy.
• Curated agricultural image datasets from multiple sources, applying manual quality filtering and hash-based duplicate detection to improve dataset quality.
• Developed preprocessing, augmentation, training, and evaluation workflows for the computer vision pipeline.

Research Intern at Suvidha Foundation (2025-06 – 2025-07)

• Developed a Transformer-based abstractive summarization pipeline comparing BART and LongT5 across news and long-form research documents.
• Evaluated model performance on CNN/DailyMail and ArxivSum datasets using ROUGE-based metrics.
• Built NLP workflows covering preprocessing, tokenization, inference, batching, and model evaluation.
• Optimized long-document experimentation under GPU constraints through efficient batching and resource-aware execution.

M.Tech in Artificial Intelligence & Data Science – K. J. Somaiya School of Engineering, Mumbai (2024 – 2026)

B.Tech in Mechanical Engineering – Government College of Engineering, Karad (2020 – 2024)