About

I'm Sia. I studied Cognitive Science and Computer Science at UC San Diego. I like working across a whole product, from deciding what to build to shipping it.

I built Chalky, a workout app I wanted for my own training. It's live on the App Store, and I own it end to end: design, development, launch, and growth.

Before that, I was a founding engineer at Uniwise, building AI tools for organ-donation teams. I worked across the full stack, mostly backend and AI, plus AWS, and talked directly with the technicians who used it. Try the interactive demo of both products.

Projects

Skills & tools

Grouped by the kind of work. Most of it comes from building Chalky and Uniwise.

AI & LLM systems

  • Prompt engineering
  • Structured outputs
  • Agentic workflows
  • OpenAI & Gemini APIs
  • Google ADK
  • Multimodal video analysis
  • OCR & document parsing
  • Vector search
  • Model evaluation
  • PyTorch

Backend & data

  • FastAPI
  • REST APIs
  • WebSockets
  • Server-Sent Events
  • JWT auth (Auth0, Clerk)
  • PostgreSQL
  • MongoDB
  • Supabase
  • Row-level security
  • Idempotency
  • S3

Mobile & frontend

  • React
  • React Native
  • SwiftUI
  • StoreKit 2
  • HealthKit
  • Offline-first sync
  • Material UI
  • HTML & CSS

Product & design

  • User interviews
  • Workflow mapping
  • Rapid prototyping
  • UI/UX design
  • Figma
  • Design systems

Cloud & delivery

  • AWS (EC2, S3)
  • Docker
  • Nginx
  • GitHub Actions
  • pytest
  • PostHog

Languages

  • Python
  • TypeScript
  • JavaScript
  • SQL
  • Swift
  • C++

Experience

Chalky

- Present

Irvine, CA

Founder & Product Engineer

Founder and engineer behind Chalky, a gym log and form-check app live on the App Store. It turns free-form workout notes into structured data, insights, and coaching, and reviews lift form from video. I own it end to end: research, product, design, engineering, launch, and support.

  • Ran a TestFlight beta with more than 60 testers, then reworked onboarding, logging language, and failure states based on where they got stuck.
  • Shipped to the App Store and handle user support directly.
  • Built the native iOS app, Supabase backend, and offline-capable sync layer.
  • Designed a coaching loop where policy code owns scoring and safeguards, while models handle interpretation and drafting.
  • Added structured outputs, provider failover, idempotency, caching, and server-side quotas so failures stay recoverable.

Worked with SwiftUI, SwiftData, Python, Supabase, Clerk, and structured LLM outputs.

Uniwise

-

La Jolla, CA

Founding Full-Stack Engineer

One of the first three engineers. I took two donation-support products from technician interviews to deployed MVPs: a live call copilot, and ContraSight, which screens donor records for contraindications. Full stack, leaning toward backend and AI.

  • Interviewed donation technicians, mapped how their calls and record reviews actually run, and decided what to build first.
  • Live call copilot: a Recall.ai bot joins the call, transcripts stream to the screen over WebSockets with speaker labels, and MongoDB keeps each call's state tied to its case.
  • ContraSight: Marker, PyMuPDF, and Tesseract read scanned records, then Gemini checks 71 contraindications. Every finding cites the exact text, highlighted in an annotated PDF.
  • FastAPI services behind both products: Auth0 JWT auth, per-organization scoping, AES-256 encrypted S3 storage, and Server-Sent Events for live progress.
  • React and TypeScript front ends, including the reviewer where each finding jumps to its source text. Deployed ContraSight with Docker and Nginx on AWS.

Worked with Python, TypeScript, FastAPI, React, WebSockets, Server-Sent Events, MongoDB, S3, Auth0, Recall.ai, Marker, Tesseract OCR, and Gemini.

Try the interactive demo of both products

UC San Diego

-

La Jolla, CA

Instructional Assistant, Cognitive Science

I helped teach two Cognitive Science courses and supported more than 100 students through labs, discussions, office hours, and project feedback.

  • COGS 150: Large Language Models & Cognitive Science
  • COGS 10: Cognitive Consequences of Technology

Covered transformers, PEFT/LoRA, prompt engineering, Python, and project feedback.