I'm a Full Stack Developer at T-Mobile, where I build the automation behind its Akamai WAF and edge-security platform: Python pipelines, React dashboards, and an LLM agent that drafts security attestations. Before that I modernized customer-facing React interfaces at Hartford Financial Services Group and built Java and Spring Boot services for USAA and Proximus at Tata Consultancy Services.

Five-plus years of full-stack work, backend-leaning, with AI shipped into production rather than pinned to a slide.

• Zero-touch request pipeline: polls a mailbox through Microsoft Graph, runs an 8-stage validation, then files the ServiceNow task and replies in-thread on its own.
• WAF attestation agent: an LLM agent on Google ADK and LiteLLM that drafts attestations, grounded in a ~1,700-line deterministic decision engine.
• Threat correlation: hourly correlation of Akamai SOCC alerts against Splunk, turning 97 alerts into 22,256 correlations with no analyst action.
• The platform underneath: a 39-service health monitor with capped auto-restart, a job runner with live logs, and Terraform-backed rollouts across 365 properties.

Question answering over 360 real SEC filings (10-K, 10-Q, 8-K) that cites the filing and page it used, or declines when the answer isn't there. Every component stayed only if a controlled experiment on the FinanceBench benchmark showed it helped.

• Held-out accuracy 15% → 49% on FinanceBench (+34 points, 95% CI); right page in the top 5 16% → 57%
• Eight experiments; BM25 hybrid and query rewriting were measured and dropped
• CI eval gate on every push, streaming API, UI, and tracing

Python Qdrant Sentence Transformers Cross-encoder rerank Ollama FastAPI Streamlit Arize Phoenix GitHub Actions

Project site · Repository

An ML forecast of South Africa's 2026 domestic T20 title. It scrapes and cross-checks five data sources, blends four models into an ensemble, and simulates the rest of the tournament 20,000 times. A React dashboard explains every number.

• 542 matches since 2011-12, merged and cross-checked across Cricbuzz, Cricsheet, ESPNcricinfo, CricTracker, and Open-Meteo
• Elo, logistic regression, random forest, and gradient boosting ensemble; 58.5% season-by-season back-test on 330 held-out matches
• Player-rated lineups, a washout model built on 15 years of venue rainfall, and Monte Carlo simulation of every stage

Python pandas scikit-learn React Vite SVG charts GitHub Actions GitHub Pages

Live dashboard · Repository

Movie ticket booking for Hyderabad cinemas, with live seat maps, snacks, and Stripe checkout, plus two Gemini tool-calling assistants: a customer concierge that finds shows and builds the order, and a manager's assistant that reports occupancy and sales and schedules shows.

• Live seat maps over WebSocket with expiring seat holds; Stripe checkout confirmed by webhook
• Gemini agents that only act after the user confirms
• JWT auth with rate limiting; Supabase Postgres via Flyway, Supabase Storage, TMDB posters

Spring Boot 3 Java 17 React Supabase Postgres Flyway Stripe API WebSocket Gemini AI JWT

Repository

• M.S. Computer Science, University of Central Missouri
• B.Tech, Electronics and Communication Engineering, Seshadri Rao Gudlavalleru Engineering College

_{My real contribution graph, regenerated daily from GitHub by a workflow in this repo.}