"It's not who I am underneath, but what I do that defines me."
| Module | Component Technologies |
|---|---|
| The Mainframe (Web Infrastructure) |
          |
| The Cyber-Deck (AI/ML Engine) |
            |
| The Sentinel (Robotics Stack) |
      |
| The Fortress (DevOps & Security) |
      |
| The Archive (Data Layer) |
       |
Status: ACTIVE | Classification: UNDERGRADUATE SPECIAL PROJECT
In complex environments, inexplicable machines are liabilities. Trust is failing.
- Tactical Solution: Deployment of the Intelligent Digital Twin system. Fusing ROS2 Humble navigation stack with Gazebo simulation and Gemini LLM to create a robot with conscience.
- Mission Parameters:
- Conversational Memory: 5-10 turn dialogue with spatial context resolution.
- Explainable AI: Real-time navigation explanations (<2s latency, >90% comprehension).
- Digital Twin Anomaly Detection: ML-based deviation monitoring with >80% accuracy.
- Tech Stack: ROS2 | Gemini-API | React-TypeScript | Gazebo | Python | Whisper | SQLite | PyTorch
Status: COMPLETED | AWARDED | Classification: BEST PROJECT - BITS TECH FEST
Gamified sustainability platform weaponizing psychology for good.
- Impact: Engaged 200+ users in real-world eco-activities through competitive digital platform.
- Tech Stack: React-Native | Gamification Engine | Sustainability Analytics.
Status: PRODUCTION | Classification: GRAPH DATABASE MANAGEMENT SYSTEM
Data fragmentation in academic research is critical. Vital connections between papers are lost in linear databases.
- Tactical Solution: Initialization of GraphLit - a Neo4j-powered system treating literature as networks, not lists.
- Performance Metrics:
- Citation Query: 6ms (vs 50ms SQL).
- Multi-Hop Chain: 24ms (vs 500ms+ SQL Recursive CTE).
- Tech Stack: Neo4j 5.26.12 | Cypher | Graph Data Science Library 2.22.0 | OpenJDK 17
Status: COMPLETED | AWARDED | Classification: HEALTHCARE AI
AI-powered prescription reminder chatbot for elderly care using Botpress and Telegram API.
- Mission: Demonstrably improved medication adherence through empathetic, timely interactions.
- Recognition: 1st Place Winner - HACK-A-BOT Hackathon 2025.
Status: RESEARCH COMPLETE | Classification: COMPUTER VISION RESEARCH
Breakthrough: 5.7x speedup on 2048x2048 satellite imagery using Vision Transformers with adaptive resolution feature extraction.
- Technical Achievements:
- Baseline (Sliding Window): 9408ms per image (81 ViT passes).
- Adaptive (Attention-Guided): 1796ms per image (5.7x faster).
- Feature Similarity: 88.4% maintained accuracy.
- Model: ViT-Large with SatMAE pretrained weights (FMoW dataset).
- Tech Stack: PyTorch | Vision Transformers | CUDA | timm | UC Merced Land Use Dataset
