Diego Jose Robles Costa

CS undergraduate at the University of South Florida building full-stack and AI software — multi-tenant web platforms, model-serving APIs, and the infrastructure behind them.

Tampa, FL · B.S. Computer Science, University of South Florida

Building software end to end

I'm a computer science student at USF who builds software end to end: a multi-tenant recruiting platform on React, TypeScript, and Postgres, backend and infrastructure work on an LLM healthcare app presented at Google DevFest, and a model-serving REST API with its own CI and secret-scanning pipeline. I work hands-on with Python, TypeScript, SQL, and C, and I care about the parts most people skip — row-level security, offline test suites, and keeping secrets out of the browser. I'm looking for a software engineering internship building AI and accelerated-computing products.

Where I've worked

Outreach Director — Society of Hispanic Professional Engineers

Sep 2026 – Present · Tampa, FL

Lead corporate outreach for USF's SHPE chapter, securing sponsorship from major technology companies to fund Hackabull, the chapter's hackathon. Manage the sponsorship pipeline from first contact through commitment while building relationships with company recruiters and engineering teams.

IT Intern — WS Development

Jul – Aug 2026 · Boston, MA

Administered 511 user identities across Windows Server, on-premises Active Directory, and Azure/Entra ID hybrid environments spanning 100+ properties and 400+ tenant companies. Scripted Group Policy deployment of Claude Co-Work to 500+ employees across Active Directory and helped build a Claude Skills library distributed through SharePoint. Remediated 193 NinjaOne-flagged vulnerabilities while working from a queue of roughly 200 weekly tickets.

Cooperative Internship, Python & Database Development — Global Career Accelerator

Aug 2026 – Present · Tampa, FL

Selected for the Global Career Accelerator, a competitive internship program through USF focused on Python and database development, including a project with Intel analyzing data for real-world business cases. Collaborate with peers and industry experts in interactive, team-based sessions to solve technical problems and deliver structured project assignments.

Backend Developer Mentorship Program — Google Developer Group

Aug – Nov 2025 · Tampa, FL

Selected as 1 of 20 students for GDG Tampa's AI/ML and cybersecurity mentorship, building a healthcare application in Python under weekly mentor code review, and continued contributing to the codebase after the program ended.

Highlight

Presenting AI Doctor App at Google DevFest

Last November at Google DevFest, we presented our AI Doctor App, the healthcare application my team and I built throughout the GDG mentorship program. Showcasing it to developers and industry professionals, and hearing their feedback on our architecture and user experience, it was an incredible experience.

The event offered more than the presentation itself. Hearing from experienced engineers and founders gave me new insight into model integration, responsible deployment, and how to position technical projects in a professional setting. I left motivated to keep improving the project.

Highlight

Completing my IT internship at WS Development

This summer I wrapped up my IT internship at WS Development in Boston, working on the identity and security infrastructure behind 100+ retail and mixed-use properties and 400+ tenant companies. I administered 511 user identities across Windows Server, on-premises Active Directory, and Azure/Entra ID, and remediated 193 NinjaOne-flagged vulnerabilities in six weeks.

Most of what I learned came from the ticket queue, roughly 200 a week of server, network, and authentication issues. Little of the work involved writing code; however, resolving authentication failures across distributed site networks gave me a formidably practical understanding of how enterprise systems hold together. Maintaining the infrastructure documentation as I went.

The project I've learned the most from

Offr+

GitHub ↗

A networking pipeline for finance students to connect with and organize their network.

Offr+ started with friends. I was the one in the code while they generated ideas for features and produced the artwork. Contacts move through custom pipeline sections, follow-ups and coffee chats land on a calendar automatically, and applications and resume versions hang off the people who matter to them. It runs locally and was never deployed, so it has no users; however, the parts I had to get right to make it work at all, row-level security across thirteen tables, OAuth refresh tokens kept server side, and a shared cache holding a quota-limited API inside its free tier, taught me more than anything else I have built.

ReactTypeScriptSupabasePostgreSQLOAuth 2.0Microsoft Graph

Projects

01 ↗

Offr+ — Networking Pipeline for Finance Students

A pipeline for moving and organizing contacts from cold outreach through to real connections. Eleven pages backed by 13 Postgres tables and 39 row-level security policies, so no session can reach another user's data. Three Supabase Edge Functions keep OAuth refresh tokens and API secrets server side while handling Microsoft Graph calendar sync, Teams meeting provisioning, and a job feed held under a 1,000-call monthly free tier by a shared 12-hour cache keyed on normalized queries. I built my skills with React, Postgres, and OAuth mostly by breaking this software apart and working out the problems. It runs locally and was never deployed.

02 ↗

AI Doctor — Healthcare Chat Application

Backend and infrastructure work on a four-person mentorship team, continued after the program ended. Fixed an auth middleware bypass that had left every endpoint unauthenticated, added rate limiting to public endpoints, enforced a clinician-only role, and moved deployment secrets to AWS SSM. Wrote the medical corpus loader and retrieval evaluation harness for the Chroma vector store, built the AWS deployment stack, and presented the architecture at Google DevFest.

03 ↗

AI-Server — Model Inference REST API

A Flask API serving a DistilBERT sentiment model behind API key authentication and per-client rate limiting, with health, predict, and model-listing endpoints. Moved model loading from import time to first use after the import chain began downloading a model, which had made the test suite impossible to run offline; added a handler injection seam so tests substitute a fake. Ships an offline pytest suite and a three-job GitHub Actions pipeline covering tests on Python 3.11 and 3.12, ruff lint, and a security job that scans full commit history for leaked secrets.

04 ↗

Email Database — In-Memory Record Store

My first project. A REPL-driven record store in C with a hand-written statement parser, fixed-layout row serialization, insert and select execution, and manual memory teardown.

Credentials

Google AI Essentials

Google · Certificate of Completion

C++ Programming Fundamentals

Microsoft · via Coursera · Aug 2026

Verify credential ↗