shirley s · b.tech cse · class of 2029
builder of things that matter
I approach problems from different angles and build solutions that have real-world impact.
Complex problems, clear solutions — that's the challenge I enjoy most.
| area | focus |
|---|---|
| 🛡️ AI Safety | real-time threat detection, behavioral analysis |
| 🎮 Systems | game loops, state logic, structured design |
| 🌐 Browser Extensions | DOM manipulation, real-time processing, APIs |
| 🧩 Problem Design | simplify → structure → ship |
AI-Powered Recruiter Cognitive Twin | Featured Hackathon Project
Problem: Traditional ATS systems rely on keyword matching, often missing highly qualified candidates.
Solution: An AI-powered hiring platform that combines semantic search, hybrid ranking, and explainable AI to identify the best candidates beyond keywords.
Features:
- Semantic resume search using FAISS
- AI-powered candidate ranking
- Hidden Gems detection
- Explainable AI insights
- Resume parsing & JD analysis
Tech Stack:
- Frontend: HTML, CSS, JavaScript (Netlify)
- Backend: FastAPI (Render)
- AI: Sentence Transformers, FAISS, OpenAI
- repository - demo - Website Link
AI Real Estate Concierge RAG + Gemini LLM Featured Project
Problem: Finding verified properties and personalized recommendations is often slow and unreliable.
Solution: An AI-powered real estate concierge that uses a verified Aqaar knowledge base to deliver grounded property recommendations and comparisons.
Features:
- Gemini LLM-powered conversational assistant
- RAG-based retrieval from verified Aqaar knowledge base
- Gemini Vision for image-based property search (implemented; subject to API quota limits)
- Intelligent intent detection and lead capture
- Property comparison using verified brochure data
- Admin dashboard with analytics and conversation insights
Tech Stack:
- Frontend: Netlify
- Backend: Node.js (Render)
- Database: Supabase
Chrome Extension AI Security Tool featured
Problem: Unsafe chats lack real-time detection of harmful behavior
Solution: Chrome extension that flags risky messages inside chats
Features:
• Real-time message risk detection
• Keyword + behavior pattern tracking
• Message highlighting with severity levels
• Live risk timeline across conversation
Web Application AI System Healthcare Logistics
Problem: Blood Type availability is delayed in emergency situations
Solution: System to track, predict, and route blood supply in real-time
Features:
• Live blood inventory tracking
• Demand prediction using past data
• Smart routing for faster delivery
• Centralized monitoring dashboard
AI Healthcare System IoT + AI featured IoT Sensors JavaScript
Problem: Lack of continuous monitoring for early health risks
Solution: IoT-based system for real-time health tracking and alerts
Features:
• Live vital monitoring (heart rate, temperature)
• Risk classification (Normal / Moderate / Critical)
• Voice-based interaction support
• Emergency alert system
Data Engineering Analytics Healthcare featured Snowflake Tableau
Problem: Hospitals lack real-time visibility to prioritize resources during emergencies
Solution: Data pipeline + analytics system that ranks hospitals and predicts demand using Snowflake SQL
Features:
• Priority scoring (0–100) for hospital risk ranking
• Critical alert detection (beds, ICU, blood shortages)
• Demand trend analysis using SQL window functions (LAG)
• Predictive alerts for rising demand (early warning system)
• Centralized dashboard combining all metrics
IoT AI Security Healthcare System featured MQTT Influx-DB Node-RED
Problem: Healthcare data is vulnerable to unauthorized access
Solution: Secure IoT system that detects threats and stops data flow
Features:
• Multi-device architecture (publisher, broker, subscriber)
• Real-time sensor data via MQTT
• Threat detection with automatic response
• Data flow control during attacks
C Team Lead Game Loops State Handling
Problem: Learning game logic lacks practical implementation
Solution: Built a full arcade game from scratch in C
Features:
• Collision detection system
• Complete game loop implementation
• Custom mechanics (Ghost Mode)
• Team-led development and integration
- Building, breaking, and fixing things
- Learning systems design from first principles
- Looking for problems worth solving
"I like simplifying complex problems into clear, practical solutions —
especially when they solve something meaningful for real people."