Monica Wu

01

About

Software that moves $200M a month doesn't get to break. That's where I learned to ship.

I'm a full-stack product engineer who's built the production surfaces — and the systems behind them. Work I've done includes high-stakes systems: an enterprise payments platform processing $200M+/month (Accenture), the component library behind healthcare apps used by physicians and pharmacies nationwide (IQVIA/Inovalon), and accessible UI deployed across ~50,000 Microsoft partner websites (Angular, React, TypeScript, Python).

When I migrated that component library across 6 major Angular versions, downstream teams saw zero regressions. That's the bar I build to.

I started in design — a BA in Art & Design — before I became an engineer. It's why I've always owned the entire surface a user touches — design systems, component libraries, interaction design. Not just the code behind it.

Now I build AI agents end-to-end: Digital Twin, a production RAG agent with structured tool-calling and real-time intent routing (Python, OpenAI, ChromaDB) — live on Hugging Face Spaces and serving real visitor queries since May 2026. Behind it, an eval harness where chunking, embedding, retrieval, and generation are each independently testable and swappable — because I think about failure modes and observability, not just whether the demo works.

I prototype and ship with AI agents in my editor daily. That started at IQVIA, where our clinical apps had to run locally on-site and client hardware varied site to site — so I tested whether LLM inference could clear that bar: Llama 3.1 on Ollama on an aging laptop as a worst-case baseline. If it ran there, it would likely run anywhere our clients deployed.

The conversations I want: product engineering teams at AI-native companies shipping LLM features to real users — where the interface and the model behind it are the same job.

10 yrs Shipping production software to real users

$200M+ Monthly transactions on a payment platform I shipped features for

RAG + agent Live in production, end-to-end

contact@wumonica.com

02

Projects

Production AI agent (RAG + structured tool-calling + real-time intent routing) that answers natural-language questions about my professional background in my own voice — GPT-4.1 Mini grounded by semantic search over ChromaDB. Shipped to production in 6 weeks; maintained independently since May 2026.

Semantic-chunking redesign

Semantic-chunking redesign improved retrieval coherence over the naive baseline — retrieval returns whole, word-aligned passages instead of mid-word fragments.

Swappable pipeline layers

Every pipeline layer — chunking, embedding, retrieval, context assembly, generation — is independently testable and swappable. Each component can be replaced without touching the rest.

Real-time agentic routing

Real-time agentic routing delivers hire/collaboration intent to a human stakeholder within seconds of a visitor interaction.

Production guardrails

Privacy and integrity guardrails: no personal contact disclosure, no fabricated opinions, no binding commitments.

How it's evaluated

Retrieval quality is verified, not assumed: before/after comparison across the chunking change on a fixed question set, guardrail behavior spot-checked against adversarial prompts, and semantic cluster separation inspected in the companion pipeline to confirm passages group coherently rather than fragmenting.

Python
OpenAI API
GPT-4.1 Mini
ChromaDB
Gradio
Hugging Face Spaces
RAG
Agentic tool-calling
Semantic chunking

Open-source Jupyter walkthrough of a full RAG pipeline — ingestion → semantic chunking → embeddings → clustering → interactive 3D visualization (demo corpus: Netflix Culture Memo).

Python
ChromaDB
Embeddings
Semantic chunking
UMAP
KMeans
Cluster eval
Vector search
Jupyter

03

Stack

AI & LLM

Large Language Models
OpenAI API
GPT-4.1 / Mini
Local / on-prem LLMs
Ollama
Llama 3.1
LLM orchestration
Fine-tuning fundamentals

Retrieval & vector

RAG pipelines
ChromaDB
Vector databases
Embeddings
Semantic search
Semantic chunking

Agents & evals

Agentic tool-calling
Function calling
Multi-turn prompt design
Structured outputs
Eval harnesses
Context-window management
Guardrails

Ship & deploy

Production AI systems
Hugging Face Spaces
Gradio
CI/CD pipelines
Model evaluation
0→1 product build

Languages

Python
Pandas
TypeScript
JavaScript
Node.js
SQL

Domains

Regulated / clinical
Healthcare data
Payments ($200M+/mo)
Stakeholder-facing delivery
Accessibility / WCAG

04

Experience

Aug 2025 — present Independent

Product Engineer — AI/LLM Systems

Shipped Digital Twin — a production RAG agent with agentic tool-calling and real-time intent routing to a human stakeholder (Python, OpenAI GPT-4.1 Mini, ChromaDB, Gradio) — live on Hugging Face Spaces. Architected the pipeline so every layer — chunking, embedding, retrieval, context assembly, generation — is independently testable and swappable, and implemented privacy and integrity guardrails: no personal contact disclosure, no fabricated opinions, no binding commitments.

Python
RAG
OpenAI API
Agentic AI
Evals

Sep 2023 — Aug 2025 Inteliquet, an IQVIA business

Software Development Engineer IV

Self-initiated feasibility testing of local LLM inference (Ollama, Llama 3.1 8B) on deliberately constrained hardware as a worst-case baseline for varied client-site environments — producing a go/no-go signal for AI features within a locked local-deployment model. Owned features end-to-end — proposal through production — for clinical research applications in a regulated healthcare environment, serving 9 client sites in tight feedback loops with engineering and clinical-ops stakeholders. Consolidated 10+ TypeScript/Angular components into a shared library and shipped i18n localization across 2 production applications, cutting duplicate implementation work across 3 product teams.

Local LLMs
Ollama
TypeScript
Healthcare data
CI/CD

Jan 2022 — Aug 2023 Inovalon

UI Engineer

Migrated an enterprise Angular component library across 6 major versions (v8 to v14) with zero regressions in the design system under healthcare apps used by physicians and pharmacies nationwide. Designed, built, and shipped the design system's core components end-to-end for an enterprise NPM library — owning everything from component specs to production code — and authored component specifications adopted as the single reference by Inovalon internal teams.

Component libraries
Zero-regression migration
Accessibility / WCAG
TypeScript

Jul 2018 — Jan 2022 Accenture

Software Engineer (Senior Analyst)

Shipped production features end-to-end for an enterprise payments platform processing $200M+/month in money movement in a regulated, high-reliability environment. Diagnosed and resolved critical production defects in payment workflows through root-cause analysis, and mentored 2 engineers through advanced Angular (RxJS, state management, lazy loading) via structured sessions and code review.

Payments ($200M+/mo)
High reliability
Root-cause analysis
TypeScript

2016 — 2018 Win-Kel · Microsoft · NBCUniversal · UIEvolution

Software / Front-End Engineer

Architected and shipped a full-stack product (React, Redux, Node.js, TypeScript) 0→1 at a 4-person startup, setting front-end technical direction and mentoring 2 junior engineers. At Microsoft, built a production Angular/TypeScript interface integrated with REST APIs for publishing and managing JSON schemas that drive automated web-page generation, and built and remediated accessible (WCAG-compliant) framework components deployed across ~50,000 partner websites. Shipped bug fixes, analytics instrumentation, and SEO sitemap improvements for social and video features on NBCNews.com — a property serving tens of millions of users monthly — and built production digital-signage web apps for smart TVs, kiosks, and mobile at UIEvolution, with performance and stress testing against real device constraints.

0→1
React
Accessibility
Mentorship

05

Education & certifications

Issued May 2026

SuperDataScience

AI Engineer Sprint — Capstone: Digital Twin, production RAG agent shipped in 6 weeks

Production-focused AI Engineer Sprint. Capstone became the first version of my Digital Twin agent — shipped to production in 6 weeks and maintained independently since May 2026: semantic chunking, privacy guardrails, and real-time agentic tool-calling.

RAG
OpenAI API
ChromaDB
Semantic chunking
Production deployment

B.A.

California State University, Los Angeles

Bachelor of Arts (B.A.) · Art — Design Concentration

Concentration in visual systems, hierarchy, and interaction design — the training behind the design systems and component specifications adopted as the single reference by consuming teams at Inovalon (10+ production components).

Systems thinking
UX fundamentals
Visual communication

06

Get in touch

If your AI feature is stuck between demo and production, message me the bottleneck — I'll reply with how I'd attack the first two weeks.

The conversations I want: product engineering teams at AI-native companies shipping LLM features to real users — where the interface and the model behind it are the same job. Fully remote, distributed-first.

Reach me directly