0 LY

▶ SOFTWARE ENGINEER

Joshua Ta

Joshua Ta

Hi! Welcome to my website :)

I'm a full-stack SWE with a background in robotics, AI development, and web development.

I'm also a huge Star Wars fan, play 5 instruments (cello, piano, guitar, drums, and ukulele), and can solve 4 different types of Rubik's cubes!

Check out some of my cool projects!

01 / PROJECTS

Things I've built

2022 → 2026
withevie.ai01
Evie booking widget at Lou's Athletic Club
FOUNDER PROJECT · ACQUIRED BY PARTYPLACE

Evie Booking Widget

An LLM-powered booking widget for event venues that I built solo. It sits on each venue's own site and takes a group from a natural-language inquiry to a paid reservation, with one flow for package, flexible, and group bookings. Guests who drop off get a single Twilio SMS with a resume link, tracked in Amplitude. Now running an iMessage channel over the same assistant that only creates bookings after Stripe webhook confirmation.

15+
venue sites
+35%
booking conversion
$15K+
processed
Next.jsTypeScriptNestJSPostgreSQLOpenAIStripeTwilio
▶ withevie.ai ▶ live at Lou's
uroassist.net02
UroAssist
CONTRACT · OC UROLOGY ASSOCIATES

UroAssist

HIPAA-compliant clinical AI assistant. Physicians ask questions in plain English and get answers grounded in the practice's own EHR: a FastAPI backend integrating ModMed via FHIR with OAuth/JWT, a Qdrant vector database doing RAG semantic search over records, and a local Llama model in Docker on AWS so patient data never leaves the practice's environment.

7
physicians
50+
patients
80%
faster lookups
ReactFastAPIQdrantLlamaDockerAWSFHIR
▶ uroassist.net login-walled (real patient data)
~/kalshi-market-sim03
$ ./bench --orders 10M book ....... price-time priority cancel ..... O(1) queue ...... MPSC → matcher gateway .... TCP, multithreaded 2.87M orders/sec ~348 ns median latency
SYSTEMS · OPEN SOURCE

Kalshi Matching Engine

Rebuilt the matching core of a prediction-market exchange in C++17 to learn how trades actually execute: price-time priority order book with limit and market orders, O(1) cancellation, an MPSC queue feeding the matcher, and a TCP gateway. Moving to integer-tick pricing and removing allocations from the hot path got it to 2.87M orders/sec at ~348 ns median latency (my machine, so treat as relative).

C++17POSIX SocketsMultithreadingLow Latency
▶ view on github
brown econ research04
LLM analyst-report pipeline
RESEARCH · BROWN ECONOMICS

LLM Analyst-Report Pipeline

LLM pipelines for Prof. Rafael La Porta's research (with Harvard's Andrei Shleifer) on analyst growth forecasts. I deployed Llama and FinBERT on Brown's Linux HPC cluster to mine 120,000+ analyst reports, and raised the LLM scoring fill rate on an earnings-call pilot from 30% to 98% by tracing a smart-quote JSON parsing failure.

120K+
analyst reports
30→98%
scoring fill rate
PythonLlamaFinBERTHPCPlaywrightStata
▶ SIDE QUESTS
Cad4Kids

Cad4Kids

Built at USC: a GUI that converts 2D drawings into interactive 3D models exportable as STL files, so kids can design something and print it. Vectorized NumPy made rendering 40% faster.

Python · PyQt5 · OpenGL · NumPy
VEX Robotics

VEX Robotics Champion

Founded and led team 2922C to the California regional championship and the top 0.3% globally. PID control and odometry motion planning doubled autonomous speed and cut positional error 60%.

▶ competition videos
GPU mining dashboard

GPU Mining Operation

My first systems project. Tuned RTX 30-series rigs to ~150 MH/s and wrote a Flask + Matplotlib dashboard tracking hashrate, temperature, and power per GPU. $7,000 in, $25,000 out over 2.5 years.

Python · Flask · Matplotlib · Hardware
Coding Campers

Coding Campers

Free coding program for elementary and middle-school students, funded by a $5,000 Dragon Kim Foundation grant. I wrote the curriculum and taught Python, Scratch, and binary.

Python · Scratch · Teaching

Client Websites

Freelance design-to-deployment builds for small organizations: responsive Next.js/React/TypeScript sites from Figma, deployed on Vercel with CI/CD, custom domains, SEO metadata, and Lighthouse tuning.

02 / EXPERIENCE

Where I've worked

Jun 2026 → now
Software Engineer Intern · PartyPlace

Full-stack at a group-booking startup with 200+ live NYC venues; own the venue dashboard and booking UI. Integrated Stripe deposits and card holds into the NestJS/PostgreSQL API, redesigned the Next.js frontend with Mapbox maps (bookings +12%), and built multi-room inventory with a PostgreSQL exclusion constraint that rejects double-bookings.

May 2025 → Jan 2026
Contract Software Engineer · OC Urology Associates

Built and shipped UroAssist end to end, from the ModMed FHIR integration to the self-hosted model.

Jan 2025 → now
Research Assistant · Brown Economics

LLM pipelines for Prof. Rafael La Porta. Currently replacing a brittle Selenium scraper with Playwright automation in which an LLM picks browser actions.

Feb 2025 → now
Technical Director · Full Stack at Brown

Promoted twice, Developer → Product Manager → Technical Director. As PM, led a 10+ person team shipping the Brown Music Review and Grad Engineering Council sites and reviewed 52 PRs. Now co-teach a bootcamp taking new members to their first full-stack app.

Jun → Aug 2024
Data Analytics Intern · USC

40% faster 2D-to-3D array transformation with vectorized MATLAB/Python; built an OpenGL GUI for CAD accessibility and STL visualization.

03 / TOOLKIT

What I use

LANGUAGES

Python, TypeScript, JavaScript, Java, C++, SQL

FRAMEWORKS

React, Next.js, React Native, Node.js, NestJS, FastAPI, Pandas

TOOLS

PostgreSQL, AWS (S3, Bedrock), Docker, Linux, Git, Jest, pytest, Stripe, Twilio, OpenAI API

ACHIEVEMENTS

★ 2nd Place, Palantir Coffee Cup Hackathon (2026) ★ USC Presidential Scholar ($150,000) ★ VEX CA Regional Champion, top 0.3% worldwide ★ Brown Sc.B. CS & Economics, GPA 3.85