Dec 2026New-grad · UCF Computer Science · 3.73

Deployed tothe problem.

I turn ambiguous requirements into deployed data, real-time, and AI systems. Open to forward-deployed, solutions, software, and data engineering roles starting December 2026.

Get in touchSee the work
SCROLL

A defense contractor. A DoD program. A paying client. Three different fronts before my degree. All shipped. The client's is still running.

Proof, in public

Contributions upstream to the libraries the industry runs on, tracked live from GitHub, open and merged alike.

0upstream PRs
0major repos
0combined ★
Li
nautilus_trader
0 ★
merged
streamlit
0 ★
mergedLi
alpaca-py
0 ★
open
scikit-learn
0 ★
open
starship
0 ★
openLi
llama_index
0 ★
open
cypress
0 ★
open
Chart.js
0 ★
open
pandas
0 ★
open
dask
0 ★
open

I build systems that keep working after I leave.

Deployments

Where someone hired or sponsored me to ship it: a defense contractor, a DoD-sponsored program, and a paying client.

01 / 03Forward-Deployed Software Engineer Intern · Jul–Aug 2026

GovSignals

A browser-backed verification harness for procurement-source data: isolated runs, screenshot evidence, and release-readiness gates across site, scraper, and application output.

  • Built and hardened the harness for a 24-ticket source-QA program
  • Shipped deterministic source selection, evidence capture, and regression gates
  • Landed two harness commits with 1,063 tests passing at handoff
Private client engagement · no public artifact
1,063 tests passed at handoffPythonBrowser QAData pipelines
02 / 03Sole engineer · paying client · 2026

Orleans Quality Cleaning

A real client with a real business and no engineering team. I owned the whole stack, from the marketing site to the phone line, and had it live in under thirty days.

  • Shipped the entire production stack from zero in under 30 days, solo
  • Built a 24/7 AI phone agent on Telnyx SIP with Llama 3.3 70B and RAG
  • Wired lead capture, email/CRM, ads, and payments into one event-driven flow
The Orleans Quality Cleaning site: Ottawa's Most Certified and Insured Cleaning Team, with the quote line and WSIB credentials.
Live in production · orleansqualitycleaning.caNext.jsTelnyx SIPLlama 3.3 70BRAGPayload CMS
03 / 03Forward-Deployed Technical Consultant · 2026

DIU · DoD Capstone

A DoD-sponsored UCF capstone translating operational asks into a traceable, side-by-side evaluation of two procurement-software vendors.

  • Normalized 52 cross-functional stakeholder requirements
  • Built interactive request-atlas and bowtie decision artifacts
  • Mapped 50 of 52 asks under IL5/CUI operating constraints
DoD-sponsored capstone · artifacts not public
52 requirements modeled across two vendorsRequirementsData modelingDecision systems

Systems I built

No brief, no client. Built because the problem was worth solving.

01 / 03Founder-engineer · production SaaS

Ringora

Multi-tenant power dialer: per-tenant row-level security, RBAC, isolated dialing sessions, low-latency WebRTC to the phone network.

  • Per-tenant row-level security + RBAC with isolated dialing sessions
  • Sub-second WebRTC connection setup over Twilio PSTN
  • Distributed call state on a live, enterprise-grade operator dashboard
The Ringora dialer: the marketing site beside a live dialing queue of leads with call scores.
Live in production · ringora.appNext.jsTypeScriptTwilioPostgreSQLWebRTC
02 / 03GPU applied-ML pipeline

ClipPro

Whisper Turbo transcription on GPU, Gemini segment scoring across four dimensions, FFmpeg compositing. A content pipeline that runs itself.

  • GPU Whisper Turbo transcription + Gemini segment scoring
  • Proprietary 4-dimension scoring engine (Hook / Flow / Value / Trend)
  • Resilient async flow with automated retries, powering a 50K+ follower network
The ClipPro pipeline, six stages left to right: Long Video, Whisper Turbo on GPU, Gemini AI, Scene Detection, FFmpeg, Ranked Clips.
Powers a 50K+ follower networkPythonWhisperGeminiFFmpegFlask
03 / 03Winner, Best Use of Google Gemini · BloomKnights · 2026

PredictIt

Look at a live game through Ray-Ban Meta glasses and compare the model's view with historical market context. The glasses stream to a desktop, Gemini identifies the sport, and local WebGPU vision tracks the players.

  • SwiftUI app on Meta's Wearables Device Access Toolkit, WebRTC straight from phone to desktop
  • Gemini 3.5 Flash classifies the sport from one compressed JPEG; a director switches streams past 70% confidence
  • Local WebGPU YOLO player tracking, read against historical model and market probability packs
PredictIt running live on a Knicks vs Celtics game: YOLO tracking boxes over the players and a 52 percent win-probability dial beside the market's 51 percent.
Winner, Best Use of Google GeminiGemini 3.5 FlashSwiftUIWebRTCWebGPU YOLOMeta Wearables

Open source is the interview you cannot fake. These contributions were reviewed and merged by the maintainers who own the code.

Merged upstream

3 merged
nautilus_trader #4466

Fix bracket tp_post_only docstring default to True

MERGED · Jul 2026 · x2
streamlit #15949

Fix RuntimeError: reentrant call inside _io.BufferedWriter on Ctrl+C

MERGED · Jul 2026

UCF computer science senior. Proven in the field: 1,063-test data-quality infrastructure, 52 defense requirements modeled, and a production client system shipped end to end in under thirty days.

B.S. CS
UCF · Dec 2026 · 3.73, Dean's List
1,063
GovSignals harness tests passed at handoff
52 asks
DIU requirements modeled across two vendors