Title Project Type Techstach
Fashion Item Classifer Convolutional Neural Network (CNN) to classify fashion items [Python 3.10, TensorFlow 2.x, Streamlit, PIL, NumPy]
  • Developed a high-performance Convolutional Neural Network (CNN) for fashion image classification that achieves 90.4% accuracy across 10 clothing categories.
  • Engineered an optimized neural architecture with three progressive convolutional blocks (32→64→64 filters), incorporating batch normalization and strategic dropout (0.25) between layers to enhance model generalization while preventing overfitting on the Fashion MNIST dataset.
  • Implemented advanced training optimizations including the Adam optimizer with exponential learning rate decay and early stopping mechanisms, balancing model complexity against performance requirements to create a computationally efficient solution suitable for real-time fashion item classification.
Kapda LLM powered marketplace [MongoDB,Express,React,Nextjs]
  • Architected a luxury resale platform using Next.js/TypeScript that aggregates thousands of designer listings across multiple marketplaces through a custom-built REST API with Redis caching and horizontal scaling capabilities, delivering a seamless shopping experience with millisecond response times even under peak load conditions.
  • Engineered a natural language search system leveraging custom-trained AI models and LLMs that interprets complex user queries (e.g., "something like what Zendaya wore to the Met Gala but more affordable") and translates them into structured search parameters, applying advanced NLP techniques and semantic analysis to understand style references, aesthetics, and design elements beyond traditional keyword matching.
  • Designed a high-performance data management infrastructure with custom APIs that seamlessly integrate with multiple e-commerce marketplaces, normalizing diverse data structures while utilizing MongoDB for flexible product cataloging and Redis caching to reduce database load by 65%, implementing intelligent cache invalidation strategies that maintain sub-100ms query response times while processing thousands of listing updates in real-time.
Archaism Studio Website [Vanilla! (Html/Css/Js)]
  • Developed a e-commerce wesbiste for a high-end fashion resale boutique operating between Paris and Tokyo. The project seamlessly blends luxury aesthetics with modern design to create an exclusive online shopping experience.
  • Engineered and implemented a comprehensive SEO strategy that yielded a 30% increase in organic search traffic across key luxury markets. Developed a targeted keyword strategy focusing on high-intent search terms which significantly improved organic product discovery.