Rama Krishna Ravilisetty

AI Engineer

Chalapathi Institute of Engineering and TechnologyBangalore, HyderabadAvailable from 2026-07-01 · internship

Open to roles

PythonSQLArtifical Intelliigence

Queue Cure – AI-Powered Real-Time Clinic Queue Management System

Neighborhood clinics still rely on paper tokens, manual patient calling, and static waiting areas. Patients often have no visibility into their queue position, estimated wait time, or when they will be called. Receptionists repeatedly answer the same questions ("How many people are ahead of me?"), reducing operational efficiency. Additionally, urgent cases can be buried in a first-come-first-served queue without any intelligent prioritization. We identified an opportunity to modernize clinic queue management by combining real-time synchronization, AI-assisted triage, and mobile-first patient Process We started by mapping the complete clinic workflow from patient registration to consultation completion. The primary goal was eliminating uncertainty for patients while reducing manual work for reception staff. Initially, we considered a polling-based architecture where patient screens would refresh every few seconds. However, this approach introduced unnecessary server load and delayed updates. We replaced it with an event-driven architecture using Supabase Realtime CDC and WebSockets, allowing every queue change to be pushed instantly to all connected devices. Next, we designed a queue state machine (WAITING → CALLED → COMPLETED / SKIPPED) to ensure predictable queue behavior and auditability. Results Queue Cure successfully transforms a traditional paper-token workflow into a real-time digital queue management system. Key outcomes achieved: • Real-time synchronization across receptionist and patient screens using WebSockets. • Sub-200ms queue update propagation through Supabase Realtime CDC. • Zero page refreshes required for patients to receive queue updates. • AI-assisted prioritization of ROUTINE, URGENT, and EMERGENCY cases. • QR-based access removes the need for app installation or patient accounts. • Multi-clinic architecture designed for future scalability. Reflection Given more time, I would conduct real-world testing with clinics, add a doctor dashboard and appointment scheduling, improve AI triage with clinician-validated models, and build advanced analytics for wait-time forecasting and operational insights. I would also perform large-scale load testing to prepare the system for multi-clinic deployment.

7 media files · queue-cure-ruddy.vercel.appView

IPL Crunch '26: What Actually Wins Matches — A Ball-by-Ball Intelligence Report

IPL discussions are driven by opinion, not evidence. Pundits claim the toss decides matches, coaches obsess over powerplay starts, and fans debate death-over specialists — but nobody had quantified which factor actually separates winners from losers across all 17 IPL seasons. The gap: 260,759 ball-by-ball deliveries existed but no structured analysis had tested these claims statistically. Process Built a two-module Python pipeline: data_engine.py for all feature engineering (phase assignment, dot ball flags, winner labels) and chart_engine.py for 14 production-grade Plotly visualisations. Applied binomial significance testing on the toss claim, computed RPO gaps per phase across all seasons, and built composite weighted scoring systems for batters and bowlers — validated against known elite performers. Results Toss win rate = exactly 50.0% across 1,095 matches (p=1.000 — not significant). Death-over RPO gap = +2.04 (worth +12.2 runs per match), positive in all 17 of 17 seasons — 100% consistency rate. V Kohli ranked #1 batter composite. YS Chahal ranked #1 bowler. Project produced 14 charts, a 5-page PDF report, and an executive dashboard — all from a reproducible two-module pipeline. Reflection With more time I would build a ball-by-ball win probability model using gradient boosting to quantify exactly when momentum shifts during a chase. I'd also add venue-specific phase analysis — the death-over advantage likely varies significantly between small grounds like Wankhede and larger ones like Eden Gardens.

7 media filesView

Trace

Indian SME factories can't answer buyers' sustainability questions — no auditable number exists. Process Defined the user pain, designed a transparent scoring formula, built mobile-first, shipped in days. Results Scores 59→91 across 25 verified batches. A4 PDF prints clean. Runs on basic Android in under 2 min. Reflection Interview factory managers first. Replace estimated thresholds with CPCB data. Add Hindi/Tamil UI.

3 media files · trace-navy-one.vercel.appView

This is Rama’s work on Wooble.

Build a profile that shows what you can do — and share it anywhere.

Build yours