Supply Chain Flow Visualization Dashboard
- Purpose: Multiscale geospatial dashboard with semantic zooming
- Tech: Vue 3, Deck.gl, Mapbox GL, D3.js, SBEB Algorithm, Web Workers, Bootstrap 5, Vite
- Outcomes: 51K+ orders rendered with semantic zoom transitions and inventory views
- Lessons: GPU rendering patterns, edge bundling, and semantic zoom storytelling
- GitHub: LogisticsDataSemantic-Zoom
RecipeLLM
- Purpose: Conversational recipe and nutrition assistant
- Tech: React, TypeScript, Flask, Google GenAI, MongoDB, PostgreSQL, Tailwind CSS
- Outcomes: Full-stack AI chat with dual database retrieval
- Lessons: Prompt engineering and intent-to-query orchestration
- GitHub: Recipellm
- Purpose: Real-time facial landmark tracking
- Tech: Python, MediaPipe, OpenCV
- Outcomes: 30+ FPS with 468 landmarks
- Lessons: Accuracy-latency tradeoffs and robust webcam inference loops
- GitHub: Face-Mesh-Mediapipe
Leaf-Disease-Detection
- Purpose: Plant disease classification
- Tech: TensorFlow, Keras, CNNs, augmentation
- Outcomes: 92% accuracy across 38 classes
- Lessons: Transfer learning and augmentation design
- GitHub: Leaf-Disease-Detection
Tumour-Detection-Model
- Purpose: Brain MRI binary classification
- Tech: PyTorch, ResNet
- Outcomes: 94% accuracy with strong sensitivity
- Lessons: Class balance and clinically aligned metrics
- GitHub: Tumour-Detection-Model
Meme-Sentiment-Analysis
- Purpose: Multi-modal sentiment classification
- Tech: Python, Transformers, CLIP, NLP
- Outcomes: 78% multi-class sentiment accuracy
- Lessons: Fusion strategies and sarcasm handling
- GitHub: Meme-Sentiment-Analysis