As a driven machine learning engineer, I focus on building models that learn, adapt, and solve meaningful problems. Explore my latest experience and projects, where I apply cutting-edge neural networks, LLMs, and AI deployment.
Hi, I am a third year Data Science Student at the University of Waterloo. I have a strong passion for Machine Learning and I firmly believe in creating data-driven, scalable, and high impact AI Applications. I am always looking for challenging and innovative ways to solve problems using various Machine Learning techniques.
For me, using ML is not just about following the recent hype around AI. I believe these applications should be built around a specific product to derive maximum benefit for the user
Worked with other intenrs to build multi-agent system to ensure company documents follow style guides.
Worked with other intenrs to build multi-agent system to ensure company documents follow style guides.
Used google-adk to built an MCP server to troubleshoot gateway.
Used Qwen Vision Language Model to make detailed body language analysis of people in a frame.
Finetunned Video Vision Transformer to classify the emotion of the person in video.
Designed pipeline for face extraction from frames
Build NER + translation pipeline for Custom document translation which was approved from production.
Engineered RAG system to generate personalized recommendations for customer insurance .
Relevant Courses: Computational Statistics and Data Analysis, Database Management,Optimization, Linear Algebra, Probability, Statistics, Calculus, Data Structures, Algorithms, Object-Oriented Programming.
Developed an Image Captioning application using Pytorch. Build and Trained an Vision Transformer model as encoder and a modified Transformer model as decoder. The model was able to generate descriptive captions for images with a bleu score of 60%
Build and Trained a vanilla transformer model from scratch capable of translating English to French using Pytorch.
Built a Text To Speech application using Pytorch. I implemented the Tacotron 2 model using pytorch. The application was able to convert text into speech with a natural-sounding voice.
Implemented a CNN-based image retrieval system by learning semantic image embeddings and using Approximate Nearest Neighbors (ANN) for efficient similarity search; implemented automated CI/CD deployment using GitHub Actions and Docker.