Selected Work
A comprehensive collection of projects spanning AI, operations, revenue optimization, and platform engineering.
Featured Projects
AI Orchestration Platform
Multi-Model LLM Governance & Routing
Platform-level thinking that enables teams to build AI features faster while maintaining cost control and reliability.
Filter by Category
🏃♂️ HYROX Coach AI
Intelligent Training Platform
🎯 The Challenge
Athletes needed training plans that adapt in real time. Existing solutions were one-size-fits-all and couldn't capture the unique demands of HYROX — let alone track custom simulations and workouts straight from the watch.
👤 My Role
Built: end-to-end from Apple Watch integration to AI training planner. Owned watch app, Kafka streaming, and LLM optimization. Reduced response latency by 65%.
🏗️ Architecture
- •Apple Watch → Kafka Streams → Feature Store → LLM Planner → Web/Mobile
- •Custom HYROX simulations with watch tracking
- •Real-time adaptation based on performance data
- •Intelligent routing and cost optimization for AI models
- •Personalized training plans based on individual progress
📊 Results
🛠️ Tech Stack
🚌 Vehicle Routing & Allocation
AI-Powered Multi-Modal Transport Optimization
🎯 The Challenge
Daily ground operations across airport ↔ hotel routes required complex routing with 45% dead kilometers, 25% service delays, and ₹2L+ daily fuel waste. Needed real-time optimization honoring time windows, vehicle constraints, and live traffic updates.
👤 My Role
Architected: LNS algorithm, live traffic integration, and vehicle pooling system. Led cross-functional team of 8 engineers. Reduced operational costs by ₹5L+ monthly.
🏗️ Architecture
- •Booking API → Normalizer → Pooler → Route Optimizer → Vehicle Allocator
- •Real-time ETA caching with 2-5 minute TTL
- •Large Neighborhood Search (LNS) destroy/repair optimization
- •Multi-modal transport with private/shared/dedicated modes
- •Live traffic integration with predictive lateness detection
📊 Results
🛠️ Tech Stack
🌊 Real-time Travel Recommendation Engine
Personalized Travel Discovery at Scale
🎯 The Challenge
Static recommendation systems had 8% conversion rate and 2-hour update lag. Built real-time system that processes 10M+ events daily and achieves 40% conversion with <100ms latency.
👤 My Role
Architected: real-time event processing pipeline, ML feature engineering, and recommendation API. Increased booking conversion by 40%.
🏗️ Architecture
- •User Events → Kafka → Feature Engineering → ML Pipeline → Recommendation API
- •Real-time user behavior tracking and analysis
- •Dynamic recommendation updates based on market conditions
- •Personalized content delivery at scale
- •A/B testing and performance optimization
📊 Results
🛠️ Tech Stack
📊 Kafka Streaming Infrastructure
Enterprise Event Processing Platform
🎯 The Challenge
Legacy batch processing had 4-hour latency and 15% data loss. Built streaming infrastructure that processes 10M+ events daily with <1s latency and 0% data loss across 15+ services.
👤 My Role
Led: multi-DC Kafka infrastructure, stream processing architecture, and monitoring systems. Achieved 99.99% uptime across 15+ services.
🏗️ Architecture
- •Multi-DC Kafka → Stream Processing → Real-time Analytics → Monitoring Dashboard
- •Distributed event processing across multiple data centers
- •Exactly-once semantics for data consistency
- •Real-time analytics and monitoring
- •Auto-scaling and fault tolerance
📊 Results
🛠️ Tech Stack
🎯 AI Orchestration Platform
Multi-Model LLM Governance & Routing
🎯 The Challenge
5+ engineering teams building AI features independently led to 3x cost variance, 40% model inconsistency, and zero governance. No centralized routing, fallback management, or compliance tracking across 8+ LLM providers.
👤 My Role
Owned: platform architecture, multi-provider routing logic, cost optimization engine, and observability dashboard. Led cross-team adoption across 5+ engineering teams.
🏗️ Architecture
- •API Gateway → Model Router → Fallback Chain → Cost Optimizer → Analytics Dashboard
- •Intelligent routing across multiple LLM providers
- •Automatic fallback mechanisms for high availability
- •Cost optimization and token budget management
- •Real-time analytics and monitoring dashboard
📊 Results
🛠️ Tech Stack
🗺️ Trip Planner (Multi-Agent + RAG)
Intelligent Travel Planning Platform
🎯 The Challenge
Users spent 2+ hours planning trips with 60% abandonment rate due to information overload. Needed intelligent system that processes 50+ data points per destination and delivers personalized recommendations in <30 seconds.
👤 My Role
Architected: multi-agent coordination system, RAG knowledge base, and real-time optimization engine. Built end-to-end from concept to production deployment.
🏗️ Architecture
- •Multi-Agent System → RAG Engine → Recommendation System → User Interface
- •Personalized trip recommendations based on user preferences
- •Multi-agent coordination for complex itinerary planning
- •RAG-powered knowledge base for destination information
- •Real-time plan optimization and adjustments
📊 Results
🛠️ Tech Stack
✅ Itinerary Validator
Automated Travel Itinerary Validation System
🎯 The Challenge
Manual itinerary validation took 45+ minutes per booking with 25% error rate. Needed automated system to validate 200+ travel constraints per itinerary and reduce processing time to <2 minutes.
👤 My Role
Built: constraint validation engine, real-time availability checker, and cost optimization algorithms. Reduced manual validation time by 80%.
🏗️ Architecture
- •Input Parser → Validation Engine → Constraint Checker → Output Formatter
- •Real-time validation of flight connections and timing
- •Hotel availability and location validation
- •Cost optimization and budget checking
- •Automated error detection and suggestions
📊 Results
🛠️ Tech Stack
💡 Itinerary Suggestor
AI-Powered Travel Recommendation Engine
🎯 The Challenge
Generic travel suggestions had 12% click-through rate and 3% booking conversion. Needed AI engine that processes 1000+ user signals and delivers personalized recommendations with 25%+ conversion improvement.
👤 My Role
Designed: ML recommendation pipeline, collaborative filtering algorithms, and real-time preference learning. Increased booking conversion by 18%.
🏗️ Architecture
- •User Profile → ML Models → Recommendation Engine → Suggestion API
- •Collaborative filtering for similar user recommendations
- •Content-based filtering using destination features
- •Real-time preference learning and adaptation
- •Multi-criteria optimization for suggestions
📊 Results
🛠️ Tech Stack
🏨 Hotel Recommendation Agent
Intelligent Hotel Selection and Booking Assistant
🎯 The Challenge
Users faced 500+ hotel options with 70% decision paralysis and 15-minute average selection time. Created intelligent agent that processes 20+ preference signals and delivers top-3 recommendations in <10 seconds.
🏗️ Architecture
- •Preference Parser → Hotel Database → Scoring Engine → Recommendation API
- •Natural language processing for preference extraction
- •Multi-factor scoring system (price, location, amenities, reviews)
- •Real-time availability and pricing integration
- •Personalized ranking based on user history
📊 Results
🛠️ Tech Stack
🤝 Supplier Negotiation Agents
Automated Supplier Communication and Negotiation System
🎯 The Challenge
Manual supplier negotiations took 2-3 weeks per contract with 30% success rate and ₹2+ Cr annual cost variance. Built intelligent agents that process 100+ negotiation parameters and complete deals in 2-3 days with 78% success rate.
🏗️ Architecture
- •Communication Engine → Negotiation Logic → Contract Generator → Analytics Dashboard
- •Multi-channel communication (email, API, chat)
- •Intelligent negotiation strategies based on market data
- •Automated contract generation and approval workflows
- •Real-time performance tracking and optimization
📊 Results
🛠️ Tech Stack
📝 Content Generation System
AI-Powered Content Creation and Management Platform
🎯 The Challenge
Manual content creation took 4-6 hours per piece with 40% revision rate and ₹50K+ monthly cost. Built AI system that generates 100+ content pieces daily with 4.6/5 quality score and 60% cost reduction.
🏗️ Architecture
- •Content Templates → AI Models → Quality Checker → Publishing Pipeline
- •Multi-format content generation (text, images, videos)
- •Brand voice consistency and tone matching
- •Automated quality assessment and human review workflows
- •Multi-language support and localization
📊 Results
🛠️ Tech Stack
🌐 Multilingual Translation System
Real-time Translation and Localization Platform
🎯 The Challenge
Global expansion required support for 25+ languages with 96% translation accuracy. Built system that processes 10K+ translations daily with <500ms response time and 45% user adoption increase.
🏗️ Architecture
- •Translation Engine → Context Analyzer → Cultural Adapter → Quality Monitor
- •Real-time translation with context preservation
- •Cultural adaptation and localization
- •Quality monitoring and human review workflows
- •Multi-language content management system
📊 Results
🛠️ Tech Stack
💬 Journey Chatbot
Intelligent Travel Assistant and Customer Support Bot
🎯 The Challenge
Customer support handled 500+ queries daily with 4-hour response time and 60% escalation rate. Created chatbot that resolves 85% of queries in <2s with 4.5/5 satisfaction score and 50% cost reduction.
🏗️ Architecture
- •NLP Engine → Knowledge Base → Response Generator → Human Handoff
- •Natural language understanding for complex queries
- •Integration with booking systems and travel data
- •Context-aware conversations and memory
- •Seamless handoff to human agents when needed
📊 Results
🛠️ Tech Stack
📧 Sold-Out Mail Processor
Automated Email Processing and Alternative Suggestion System
🎯 The Challenge
Sold-out notifications were manual with 2-hour delay and 40% customer churn. Built automated system that processes 200+ emails daily and suggests alternatives in <5 minutes, reducing churn to 20%.
🏗️ Architecture
- •Email Parser → Alternative Finder → Notification System → Analytics Dashboard
- •Automated email parsing and content extraction
- •Intelligent alternative suggestion based on preferences
- •Multi-channel notification delivery
- •Performance tracking and optimization
📊 Results
🛠️ Tech Stack
🗄️ Vector DB Framework
Scalable Vector Database Infrastructure for AI Applications
🎯 The Challenge
AI applications needed vector storage for 10M+ embeddings with <100ms search latency. Built framework that handles 1M+ queries daily with 99.9% accuracy and 99.99% uptime across 5+ services.
🏗️ Architecture
- •Vector Storage → Index Engine → Search API → Real-time Updates
- •Distributed vector storage with horizontal scaling
- •Optimized similarity search with sub-second response times
- •Real-time vector updates and synchronization
- •Multi-tenant support with isolation
📊 Results
🛠️ Tech Stack
⚡ API Rate Limiter
Intelligent API Rate Limiting and Traffic Management System
🎯 The Challenge
API abuse caused 40% service degradation with 2M+ requests daily. Built rate limiting system that maintains 99.9% API stability with <50ms response time and 95% abuse prevention.
🏗️ Architecture
- •Traffic Monitor → Rate Limiter → Adaptive Controller → Analytics Dashboard
- •Multi-algorithm rate limiting (token bucket, sliding window)
- •Adaptive rate adjustment based on system load
- •Real-time traffic monitoring and alerting
- •Per-user and per-endpoint rate limiting
📊 Results
🛠️ Tech Stack
📊 Group Dashboard
Real-time Analytics and Monitoring Dashboard for Teams
🎯 The Challenge
Teams lacked real-time visibility with 2-hour data lag and 30% decision delay. Built dashboard that processes 1M+ data points daily with <1s freshness and 90% user adoption.
🏗️ Architecture
- •Data Collectors → Processing Engine → Dashboard API → Real-time Updates
- •Real-time data collection from multiple sources
- •Customizable widgets and dashboard layouts
- •Role-based access control and permissions
- •Mobile-responsive design with offline support
📊 Results
🛠️ Tech Stack
💰 Price Match System
Automated Price Monitoring and Matching Platform
🎯 The Challenge
Manual price monitoring took 4+ hours daily with 15% accuracy and ₹50L+ annual revenue loss. Built automated system that monitors 100+ competitors with 99% accuracy and <5min response time.
🏗️ Architecture
- •Price Scraper → Competitor Monitor → Pricing Engine → Notification System
- •Automated price scraping from multiple sources
- •Intelligent price comparison and analysis
- •Automated pricing strategy adjustments
- •Real-time alerts and notifications
📊 Results
🛠️ Tech Stack
🔧 TME (Travel Management Engine)
Comprehensive Travel Management and Optimization Platform
🎯 The Challenge
Travel management was fragmented across 8+ systems with 3-hour data sync delays and 25% booking errors. Built unified platform that manages 10K+ bookings monthly with 100% integration and 4.8/5 satisfaction.
🏗️ Architecture
- •Planning Engine → Booking System → Travel Tracker → Analytics Platform
- •End-to-end travel lifecycle management
- •Integrated booking and payment processing
- •Real-time travel tracking and updates
- •Comprehensive analytics and reporting
📊 Results
🛠️ Tech Stack
Showing 19 of 19 projects