About Me
Aspiring AI and Machine Learning Engineer with a strong foundation in Computer Science and Engineering. Currently pursuing a Bachelor of Technology at Bennett University with a CGPA of 8.17. Passionate about leveraging deep learning, computer vision, and innovative technologies to solve real-world problems in healthcare, legal systems, and beyond. Skilled in Python, PyTorch, TensorFlow, and more, with a focus on research-driven solutions and efficient model optimization.
Key Skills
Experience
Machine Learning and Deep Learning Training and Internship
AIQUANTUM Smart Solutions Pvt. Ltd. | May 2024 – July 2024 | NITK Surathkal (Remote)
- Designed and optimized deep learning models for liver cancer detection and other computer vision tasks (medical imaging, satellite data), leveraging advanced preprocessing and architecture tuning.
- Achieved 97.50% accuracy on histopathology image classification using a custom ResNet-50, setting a new dataset benchmark.
Projects
SafeHeal
An AI-powered system for wound segmentation and classification using hybrid UNet-EdgeNeXt and Custom EdgeNeXt-Small models. Achieved 84.85% Dice and 80.42% IoU in segmentation; optimized by hybrid architecture combining spatial and contextual details. Reached 99.14% accuracy, 99 AUC, and 99.05 MCC for wound type classification across 10 classes using a custom EdgeNeXt-S model.
GitHubLawSage.AI
AI-powered legal assistant for Indian law offering case summarization, query resolution, and statute interpretation. Leverages fine-tuned, quantized LLMs (GPT-2, LLaMA, BERT-BART, Phi-3.5-mini-instruct) for faster legal text processing. Secure, interactive platform enabling authentication, chat history management, and real-time document analysis for legal professionals.
GitHubHemoVeda
A real-time anaemia detection system that analyzes eye, palm, fingernail images, and blood reports, achieving 96.06% accuracy. Handles class imbalance with CycleGAN, improves interpretability using Grad-CAM visualizations. Enables mass anemia screening via telemedicine, aligned with GOI’s Anaemia Mukt Bharat initiative.
GitHubMorphe
Detects face-swap deepfakes in real-time with 92.6% classification accuracy. Integrates custom DenseNet-121 with Coordinate Attention and multi-branch RGB, edge, texture inputs. Uses Grad-CAM to localize manipulations, aiding forensic analysis and fact-checking workflows.
GitHubResearch & Publications
- DisasterFormer: An Enhanced Transformer with Edge-Guided Attention for Disaster Image Recognition - Published at IATMSI-2025 (Conference Paper) | DOI
- TriVerseNet: Multi-Branch ResNet Framework for Enhanced Endoscopic Lesion Classification via Edge and Texture Integration - Published at NETCRYPT-25 (Conference Paper) | DOI
- A Small-Scale ConvNeXt Variant Optimized for Pneumonia Classification on Chest X-Ray Images – Accepted and Presented at IACIS-2025 (Conference Paper)
Achievements & Certifications
Certifications
- Fundamentals of Deep Learning – Nvidia
- TensorFlow Developer – DeepLearning.AI
- Exploratory Data Analysis for Machine Learning (Honors) – IBM
- Data Structures & Algorithmic Toolbox – University of California San Diego
Achievements
- Participated in HackCBS 7.0, with a project recognized for innovation and selected among the best in the Vultr Title Sponsor Track, demonstrating technical excellence and impactful problem-solving.
- Contributed significantly to a team's Top 50 finish in the HackWars Hackathon at Chandigarh University, competing against 1500+ teams nationwide and advancing to the offline showcase round.
- Solved 100+ competitive programming based question across multiple platforms like LeetCode, HackerRank, GFG.