Hi, I'm Vedant Agnihotri
B.Tech Computer Science Graduate | Machine Learning & Computer Vision Enthusiast
Final-year Computer Science Engineering student at SASTRA University, Tamil Nadu, passionate about building intelligent systems that solve real-world problems in healthcare and agriculture.
About Me
My journey from curiosity to expertise in machine learning and computer vision
My Journey in AI and Computer Vision
Picture this: a curious student in Tamil Nadu, tinkering with code late into the night, suddenly realizing that technology could bridge gaps in human communication. That's where my passion for machine learning truly sparked—as a fresh Computer Science graduate from SASTRA University (class of June 2025), my world changed during my internship at Elevate Labs in Bangalore. There, I dove headfirst into building a real-time American Sign Language gesture recognition system using MediaPipe and CNN architectures. Watching the app translate hand movements into text in real-time wasn't just a technical win; it revealed the profound impact of computer vision in making technology accessible to everyone, from the hearing impaired to everyday users.
This breakthrough moment fueled my drive for more. Eager to tackle real-world challenges, I jumped into advanced research at IIT Ropar, where I'm currently leading an agricultural computer vision project on zucchini leaf instance segmentation. By crafting an end-to-end data pipeline with Roboflow and Albumentations, and fine-tuning YOLOv8, I've pushed the model to achieve 96% mAP at IoU 0.50 and 84% for mAP50-95—metrics that prove how AI can revolutionize crop monitoring for farmers facing unpredictable challenges. Balancing this with my studies wasn't easy, but my commitment to continuous learning and sharp time management skills turned late-night debugging sessions into triumphs of academic and research excellence.
At my core, I believe in crafting intelligent systems that make a tangible difference—whether it's aiding doctors in diagnosing respiratory conditions through a PyTorch-based classification model with 81.6% accuracy, or empowering farmers with precise crop insights.
AI Research
Deep learning architectures for medical diagnosis and agricultural applications
Computer Vision
Real-time systems for object detection, segmentation, and gesture recognition
Data Science
Statistical analysis and machine learning model optimization
Software Development
Full-stack development with focus on ML deployment
Experience
Professional journey and research positions
Summer Research Intern
IIT Ropar
Working on advanced computer vision research focusing on instance segmentation for agricultural applications
- Developing YOLOv8-based models for crop monitoring
- Implementing COCO-standard annotation pipelines
- Collaborating with PhD researchers on agricultural AI solutions
Machine Learning Intern
Elevate Labs
Developed real-time computer vision applications for accessibility, focusing on ASL gesture recognition
- Architected real-time ASL gesture-to-text conversion system
- Implemented MediaPipe-based hand tracking pipeline
- Achieved cross-platform compatibility with TensorFlow deployment
Technical Skills
Comprehensive overview of my technical expertise
MY SKILLS
Programming Languages
ML/AI Frameworks
Data Science Tools
Web Development
Developer Tools
Areas of Expertise
Featured Projects
Showcasing innovative solutions in ML and computer vision
Respiratory Sound Classification System
Advanced deep learning system for medical diagnosis using spectrogram-based audio analysis
Key Achievements
- Achieved 81.6% accuracy, 76.25% sensitivity, and 92.32% specificity
- Outperformed baseline models by 5.21%
- Surpassed published benchmarks with CNN-LSTM hybrid architecture
- Multi-class classification of four respiratory sound categories
Technologies
Real-Time ASL Gesture Recognition
Computer vision application for American Sign Language gesture-to-text conversion with cross-platform compatibility
Key Achievements
- Real-time hand tracking and gesture recognition
- Custom CNN model with 3 convolutional layers
- Integrated text-to-speech functionality
- Cross-platform compatibility achieved
Technologies
Zucchini Leaf Instance Segmentation
Multi-condition agricultural computer vision system with comprehensive deep learning pipeline for crop monitoring
Key Achievements
- 96.53% mAP@0.5 detection accuracy (13-26% above industry standard)
- 95.71% segmentation quality with 87.94% localization precision
- 50-75% faster training efficiency (<50 epochs vs 100-200 standard)
- Robust performance across varying environmental conditions
- COCO format standardization with comprehensive annotation pipeline
Technologies
Research Publications
Contributing to the advancement of medical AI
Respiratory Sound Classification using Deep Learning Approaches
A comprehensive study on using CNN-LSTM architectures with spectrogram analysis for automated respiratory disease diagnosis. Our hybrid model achieved 81.6% accuracy and 92.32% specificity, outperforming baseline models by 5.21% and surpassing published benchmarks.
Get in Touch
Let's discuss your next AI project or research collaboration
Send me a message
I'd love to hear about your project or opportunity
Let's Connect
Download Resume
Get a detailed overview of my experience, skills, and achievements in machine learning and computer vision.