About Me

I'm Sia, a software engineer studying Cognitive Science + CS at UC San Diego. I build products and AI systems across iOS, backend infrastructure, and applied machine learning.

Right now I'm shipping Chalky, a fitness app for natural workout logging and personalized coaching. At UniWise, I worked across product, infrastructure, and AI systems that helped technicians move 50x faster on contraindication checks.

FounderBuilding Chalky end to end
Production AIHigh-stakes, reviewable workflows
Human-centeredCS + cognitive science

I'm most useful when a problem crosses boundaries. I collaborate closely, debug methodically, and care about getting the right answer rather than the fastest-looking one.

My Philosophy

I like owning the whole problem. I am willing to move between the interface, API, data model, infrastructure, and AI pipeline when that is what the work needs. I do not expect to know every answer up front, but I take responsibility for finding it and carrying the work through.

I care about the details that make software dependable. Clear interfaces, sensible APIs, tests, failure states, logs, and documentation are not glamorous, but they matter. I would rather fix the underlying issue than make a fragile patch look finished.

I am hungry to get better. I seek out hard problems, ask questions early, and study the parts of a system I do not understand yet. Feedback does not threaten me; it gives me a faster path to stronger work.

I want the result to be useful. I pay attention to how people actually use what I build, then adjust based on evidence. The goal is not to sound clever or use the newest tool. It is to ship something trustworthy that solves the right problem.

Projects

Technical Skills & Toolkit

ML & AI

Supervised & Unsupervised Learning
Reinforcement Learning & Optimization (PPO, genetic algorithms)
Transformer Models & LLM Fine-Tuning (PEFT/LoRA)
Prompt Engineering & LLM Orchestration
Model Evaluation & Deployment
Vector Search & RAG
LLM Security: Prompt Injection, Data Leakage, Guardrails
OCR / PDF Pipelines

Software Development

APIs: FastAPI (REST), WebSockets (real-time)
Auth & Identity: Clerk, JWT, Sessions
Supabase: Postgres, Auth, Edge Functions
Databases: PostgreSQL, MongoDB (Beanie)
Testing: pytest, Type Hints
API Design for Scale: Rate Limits, Idempotency, Audit Logs

Cognitive Science & Product

User-Centered HCI
Conversation Design
User Research & Usability Testing
PRDs, MVP Scoping, Roadmapping
Stakeholder Communication (Eng ↔ Product)
Experimentation Mindset: A/B Tests, Metrics, Funnels
Writing: Clear Specs, Decision Docs, Release Notes
Figma for UX/UI Prototyping
Responsible AI (privacy/compliance-aware product design)

Tools & Data

SQL & Pandas
Supabase RLS & Policy Design
Docker & CI/CD (GitHub Actions)
Observability: Logging, Tracing Basics
Security Basics: Secrets Management, Least Privilege
Cross-Functional Collaboration

Experience

Founder & Software Engineer

Chalky - Present

Founder and engineer behind Chalky, an AI-powered fitness app that turns free-form workout notes into structured data, insights, and lightweight coaching. Designed as a product-first system focused on clarity, low friction, and real-world usage.

  • Built the end-to-end pipeline to parse unstructured workout logs into validated training data, summaries, and weekly insights.
  • Designed LLM prompts, schemas, and guardrails for explainable, user-trustable outputs.
  • Implemented authentication, storage, and policy controls with privacy, cost, and scale in mind.
  • Led UX, pricing, and feature-scope decisions using real user behavior and feedback.
SwiftUIPythonLLM APIsStructured Outputs & JSON SchemasSupabaseClerkPrompt EngineeringProduct Design

Founding Software Engineer, Backend & AI

UniWise AI - PresentSan Diego, CA

Founding engineer on UniWise's core platform, building AI systems for organ-donation workflows. Worked across two flagship products: a real-time call-center triage tool and ContraSight, a document screening and decision-support system.

  • Built backend services for live call ingestion, transcription, and structured case generation.
  • Developed ContraSight's document pipeline to process records, extract clinical signals, and surface contraindications for technician review.
  • Designed multi-agent workflows with human review, confidence checks, and audit logs.
  • Implemented retrieval grounded in clinical policies and SOPs to improve explanations and reduce hallucinations.
PythonFastAPIWebSocketsMongoDBPostgreSQLOCR & PDF ProcessingLLM OrchestrationRAG / GraphRAGAudit Logging & ObservabilityHIPAA-Aligned Design

Course Designer & Instructional Assistant

UC San Diego - San Diego, CA

Co-designed and supported two Cognitive Science courses, combining co-teaching discussion sections, curriculum work, and student mentorship.

COGS 150: Large Language Models & Cognitive ScienceFacilitated discussions and office hours, then graded coding labs spanning ML fundamentals, AI ethics, and multimodal cognition.
COGS 10: Cognitive Consequences of TechnologyLed discussion sections for 100+ students and mentored work across research, product design, writing, and presentation.
PythonJupyterGitLLMsProduct DesignUX