Garv Kanojia

Computer Science undergraduate with expertise in software development and machine learning. Proven experience in creating end-to-end AI applications and data-driven systems, achieving significant project outcomes. Finalist in national hackathons and adept at solving over 650 algorithmic challenges. Committed to utilizing technology for intelligent, user-focused solutions.

• Multi-Class Chest X-ray Disease Classification
• Built a deep learning model to classify 14 chest diseases using 80,000 X-ray images.
• Fine-tuned DenseNet161, tackling class imbalance via synthetic augmentation and preprocessing.
• Achieved 88.67% cumulative accuracy with AUC-ROC scores ranging from 79% to 94%.
• Engineered a multi-label prediction pipeline with per-class confidence scoring.

• Financial Document Analyzer and QnA Bot
• Built a tool to extract and summarize financial PDFs (e.g., ITRs, bank statements) and answer context-aware queries.
• Designed a custom prompt pipeline combining business context and extracted data for Mistral-7B inference.
• Achieved 90% accuracy for information extraction, and 92.4% accuracy for tabular data parsing.
• Enhanced LLM performance via prompt engineering and modular summarization logic.

• Electronic Health Record Summarizer and Advisor (QnA Bot)
• Developed a system to ingest and analyze EHRs and prescriptions (PDFs and scanned images).
• Used Tesseract for OCR, and integrated LangChain with vector DBs to enable fast, memory-augmented retrieval.
• Applied PEFT and RAG on Mistral-7B for real-time question answering with high context retention.
• Delivered an AI advisor offering personalized insights from medical documents.

• Solved 650+ DSA problems on LeetCode, showcasing strong problem-solving and algorithmic skills.
• Participated in 6 hackathons; finalist at Hackearth and Hathor, building impactful AI tools under pressure.
• Attended Training on Generative AI from AI-Shala.
• Completed DSA with Java course from Apna College