Prajay Reddy Minuka

Software Engineer — GroundGame.Health

Production AI in regulated healthcare. I own systems end to end, from the event pipeline to the agents built on top of it.

• Architected a high-scale, event-driven messaging platform using Azure Service Bus, Event Hubs, and .NET Azure Functions, processing 10M+ messaging events across SMS and email channels with reliable, scalable, and fault-tolerant event handling.
• Built an agentic compliance-auditing service that pulls call and SMS transcripts through tool calls, scores each interaction against SLA and HIPAA rubrics with cited transcript evidence, and escalates violations for human review, expanding QA coverage from manual sampling to 100% of interactions.
• Launched a RAG benefits assistant for members and care staff, using Azure AI Search hybrid retrieval over payer benefit documents and citation-grounded Claude responses, resolving 65% of benefits inquiries without human escalation.
• Shipped voice AI agents on Azure AI Foundry and Twilio Voice with real-time speech and tool calls into member APIs, automating outreach calls, HIPAA-compliant identity verification, and benefits workflows with warm handoff to live agents.
• Partnered directly with payer clients to launch 17 Medicaid and Medicare member-engagement programs as production n8n workflows and .NET APIs, orchestrating multi-channel SMS, email, and call outreach across 5 health plans.

Graduate Research Assistant — University of South Florida

Research on the privacy side of LLMs.

• Ran experiments with EW-Tune, a differentially private LLM fine-tuning framework, measuring the tradeoff between privacy guarantees, training noise, and model performance.

Founding Engineer — Ragnova

Zero to one: founding engineer on an AI product, from architecture to a working MVP.

• Built an agentic website-builder MVP on the Claude API that interviews non-technical business owners, plans site structure, writes and deploys code autonomously, and ships live sites in under five minutes.
• Designed a planner–builder–reviewer multi-agent architecture with a self-correcting feedback loop and iteration caps to bound cost and latency.
• Engineered the agent tool-use layer with runtime tool discovery, JSON-schema contracts, and sandboxed file operations, enabling multi-step tool chaining, automatic error recovery, and live previews.

Software Engineer 2 — Deloitte & Touche LLP

Promoted from Software Engineer (Aug 2021 – May 2023) to Software Engineer 2 (May 2023 – Jul 2023)

Where I learned to build systems that have to be right: financial data, audit-grade reconciliation, and datasets with 50M+ rows.

• Developed backend services and RESTful APIs in C# and ASP.NET Core for Omnia DNAV, an investment management audit platform used by auditors to analyze fund holdings, pricing, and NAV.
• Automated 14 recurring audit routines as scheduled background jobs, processing 2M+ fund records across complex master-feeder and multi-class fund structures, saving 40 hours per engagement.
• Integrated third-party market-data APIs (Bloomberg, Refinitiv) to reconcile fund pricing, holdings, and security reference data against independent sources, automatically flagging discrepancies above materiality thresholds for auditor review.
• Optimized SQL Server data access with Entity Framework Core and stored procedures, redesigning indexes and queries to reduce report generation time by 35% on 50M+ row datasets.

Machine Learning Intern — National Instruments

• Trained and evaluated machine learning models on historical system data to detect and predict faults, supporting fault-tolerant system design.

fitbud — Social Weight-Tracking App

Next.js, TypeScript, Supabase, PostgreSQL, Docker

• Built a full-stack social fitness app with Next.js App Router and Server Actions handling all data access, backed by Supabase Postgres with Row Level Security enforcing every authorization rule at the database layer instead of in application code.
• Designed a friends-and-feed system — requests, likes, and comments — using PostgREST relationship embedding for nested queries and normalized foreign-key constraints to prevent duplicate mirrored friend requests.
• Containerized the app with a multi-stage Docker build using Next.js standalone output for minimal production images, deployed with CI-style auto-deploy on every push to main.

ResumeAI — AI-Powered Resume Tailoring Platform

FastAPI, React, Azure, Claude API, Docker

• Built an end-to-end platform that scrapes job postings with Playwright, tailors LaTeX resumes with Claude Sonnet 4.5 via the Anthropic API with prompt caching, and compiles ATS-optimized PDFs, cutting resume customization from hours to under 60 seconds.
• Architected a FastAPI backend with JWT authentication and an Azure SQL database, containerized with Docker and deployed on Azure Container Apps with CI/CD.
• Shipped a React + Vite PWA frontend with multi-format downloads and a companion Chrome extension for one-click tailoring from any job posting tab.