Krish Bhatia

Bus Prediction Algorithm: Designed and delivered a BLE-based algorithm that improved bus number accuracy from 70% to 95%, enhancing data integrity across auditing, reporting, and fare systems. Collaborated with the ETIM team and ground managers enabling downstream enhancements and completing end-to-end product delivery. Route Deviation Logic: Created a math-based algorithm to terminate live tracking sessions when buses deviated from assigned paths, reducing incorrect tracking events around 1000 to 50 per month and improving live tracking user ratings from 3.5 to 3.8. Nearby Bus Logic: Engineered an inter-city bus discovery system using Uber H3 indexing and Redis Geo-Hash, reducing search latency (200ms -> 27ms) and scaling load capacity from 0.1M to 0.5M concurrent requests under stress testing.

ETA Prediction Model: Developed an XGBoost-based ETA prediction model using GPS and stop-arrival data, improving accuracy from 50% to 85% and increasing user engagement by 89%. Delivered actionable insights for operations team via the Operator App. (Python, Decision Trees, XGBoost, AWS SageMaker, hyperparameter tuning) Route Prediction Algorithm: Independently developed a real-time bus movement-based prediction algorithm, increasing live bus availability by 78% and streamlining field operations. GPS Ingestion Optimization: Replaced a legacy API-based ingestion pipeline with a RabbitMQ-based architecture, reducing latency by 200ms per packet, cutting infrastructure costs and enabling scalable onboarding of new agencies. Reports Orchestrator: Designed and scaled a reporting service into the Operator App using Java, Python, ELK and Postgres, automating scheduling workflows and reducing issue triage time by 40% through cross-functional collaboration. Migration & Observability: Led ELK stack upgrade (6.7 to 8.9) and mentored junior developers while integrating Log4j2 (distributed tracing) and Datadog to improve logging and monitoring across all microservices.