🚀 Vinayak Prem Bhatia

📧 Email: [email protected]
📱 Mobile: +91 9930679651
🔗 LinkedIn: vinayak-bhatia
💻 GitHub: vvinayakkk

---

🏫 Education

• SP Jain Institute of Management & Research, Mumbai, India
  Minors in Management (Feb. 2024 - Present)

• Sardar Patel Institute of Technology, Mumbai, India
  B.Tech in CSE with Specialization in AI-ML (CGPA: 9.34, Nov. 2022 - Present)

---

💼 Research Experience

1. Research Assistant under Prof. Vaishnavee Rathod

Enhanced Image Processing for Vehicle and Crack Detection (July 2024 - Present)

• Developed a Vehicle Detection model using Vision Transformer (ViT), achieving 92.8% accuracy on 16,185 images.
• Implemented a Crack Detection system leveraging YOLOv9 and ViT, reducing false positives and improving precision for structural safety assessments.

2. Gen-AI Intern at Sardar Patel Institute of Technology

(August 2024 - January 2025)

• Fine-tuned LLMs (Gemini 2B, Gemma2B, LLaMA 3.2, Mistra-7B) and evaluated models using ROUGE, BERT, and BLEU metrics.
• Built a dynamic AI pipeline with real-time tracking and personalized feedback for improved interview performance.
• Enhanced topic coverage and introduced dynamic difficulty adjustment for personalized interview assessments.

3. Research Intern at IIT Patna

Crop Disease Detection Using ViT (January 2025 - July 2025)

• Led the development of a Crop Disease Detection system, analyzing 50,000+ agricultural images.
• Achieved 98.3% detection accuracy using advanced models like ViT, EfficientNet, and YOLOv8.
• Designed a comparative performance analyzer integrating ResNet, DenseNet, MobileNet, and InceptionNet.

4. Research Intern at SPJIMR

Research Assistant Platform Development (October 2024 - Present)

• Developed a Research Assistant Platform using RAG-based retrieval and FAISS for efficient document querying, aligning with SPJIMR’s research goals.
• Designed tools for research journey mapping, corpus analysis, and project documentation to enhance workflows.
• Built a user-friendly interface with context-aware suggestions, improving accessibility for non-technical users.

---

💡 Projects

1. Advanced Image Segmentation Models

• UNet for Cell Nuclei Segmentation: Achieved improved segmentation accuracy through advanced preprocessing techniques.
• EffUNet: Integrated EfficientNetV2 with UNet, achieving IoU scores of 0.83 (buildings) and 0.91 (roads).

2. AI-Powered Innovations

• Sandalwood Knowledge System: Built an ASR and QA pipeline with RAG architecture, optimized for Kannada speech recognition and real-time translation.
• Document Classification System(Classifi
  hackathon): Spearheaded the development of a domain-specific document classification system fine-tuned on BERT for legal and financial documents. The system achieved an accuracy of 93%, outperforming traditional machine learning models
• Graph-Based QA Platform: Integrated Neo4j and Pinecone for semantic search and real-time collaboration.

3. Featured Repositories

• Qwen2VL Image-Based OCR Query System: Developed a local image-to-text query system using Qwen2VL, optimized for systems with 16GB RAM.
• Gesture Recognition System: Built a MediaPipe-based system for detecting hand poses and gestures using TFLite models and custom datasets.
• SQL Query Assistant: LangChain-based tool for converting natural language into SQL queries, integrated with PostgreSQL, MongoDB, and Gemini AI.
• Emotion Recognition: Built an RNN to classify tweets into six emotional categories, leveraging LSTM layers and early stopping techniques.

---

🏆 Achievements

• 🥇 1st Place: IIIT Nagpur Genathon 2.0 (National Level) for innovative AI solutions.
• 🥈 Runners-Up: AIQuest Hackathon, IIT Bombay Techfest '24.
• 🥉 3rd Place: Wall Street Analytics Challenge, BITS Pilani Hyderabad '24.
• 🌍 World Rank 6: Zelestra X AWS ML Ascend Challenge.
• 🏅 Finalist: Smart India Hackathon for Alumni Connect.
• 🥉 3rd Place: MLFiesta Hackathon by IIIT Bangalore.
• 🥈 Runners-Up: Classifi Hackathon, IIT Bombay Techfest '24.
• 🏆 5th Rank: Technovate Hackathon, RC Club Mumbai.
• 🏆 6th Rank: VCET Hackathon '24.
• 🏆 AIR 201: Amazon ML Challenge 2024.

---

📜 Certifications

• Machine Learning A-Z | Udemy
• Artificial Intelligence A-Z | Udemy

---

🎯 Focus Areas: GenAI, Vision Transformers, Semantic Search, RAG Architectures, Image Processing, and Efficient AI Systems.