日月神教
Sun Moon Divine Cult
Bill Zhang
AI engineer. Builder of voice-first systems.
Loading the sect
Slow connection?
Read as page日月神教
Sun Moon Divine Cult
Bill Zhang
AI engineer. Builder of voice-first systems.
Loading the sect
Slow connection?
Read as page日月神教Sun Moon Divine Cult
AI engineer. Builder of voice-first systems.
I'm Bill Zhang. By day I build LLM agents at Pinterest. By night I build software you can talk to, which is how I ended up with 36 hackathon wins. This page is one of those builds. Go on, ask it something.
This site is a 3D flight through 日月神教, the Sun Moon Divine Cult, with a voice guide you can talk to. Both need JavaScript, so here is the whole portfolio as text.
Latency, interrupted sentences, tool calls that half-fail, a model that answers the wrong question with total confidence. At Scale AI I shipped a multi-agent system that works through up to 20K denied claims a week, and owned the eval framework that moved audit accuracy from 64% to 88%. Almost all of that came out of the failure cases.
A 911 console that triages live calls, a whole bank that runs inside a phone call, a food-rescue network that dispatches by voice, game tape you can question mid-replay. Several of them need nothing but a phone. No app, no internet.
At a hackathon I pick one idea and make it actually work. Dispatch AI was a phone line, a fine-tuned triage model, and a dispatcher screen. It won the $25K SkyDeck grand prize at Berkeley against 900+ people, plus a separate $25K investment from AIC.
Most people can speak several times faster than they type, with a fraction of the effort. And while you talk, your hands are free for whatever else you were doing. That is why this site listens instead of making you click, and why TalkTuahBank handles balances, transfers, and bill pay by voice alone. No app, no data plan.
Always-on public safety AI. Monitors building audio, detects panic, and triggers a multi-agent response in real time.
3D AI replay for coaches, analysts, and refs. Turns 2D sports footage into voice-queryable tactical views.
Empathetic 911 dispatcher co-pilot. Aggregates emergency calls, triages severity, extracts context, and keeps humans in control.
Voice banking over a regular phone call. No internet, smartphone, or banking app required.
Lectures that talk back. A voice tutor with Gemini-driven slides and emotion-aware pacing.
Real-time UCSC loop bus tracker. Helped students see live campus bus locations and reduce pressure on metro routes.
Conversational AI for bank applications. Answer the questions on a call and the form fills itself in as you talk.
Phone-call voice guidance for complex websites. A caller gets step-by-step spoken help while the page highlights what to do next.
Decentralized food rescue agents. Matches surplus suppliers to food banks and shelters with real-time logistics planning.
Lightweight browser security companion. Explains risky permissions in plain English instead of scaring users.
Transform a calming thought into an immersive VR journey with generative media and Gaussian splats.
Rubber-duck your study materials. Upload messy notes and talk through them until they click.
Voice-based food distribution assistant. Any phone can log surplus food or find nearby help.
Gaussian Splatting plus NFTs. Turn personal video into interactive 3D assets minted on Solana.
Wearable-powered social assistant. Ray-Ban live video becomes real-time bios and pickup lines.
36 hackathon wins.
DispatchAI
TalkTuahBank
SentinelAI
linguify
Tidbits
AdaptED
WebWeaver
SoundSearch
Vocalyze
Team Food Tactics
SlugMeditate
SplatNFT
PyPointer
ABSeas
XPlore
SwarmAID
Mad Lyrics
Doggo AI
Talking Terry
SlugLoop
Covinet
GitPT
IntelliConverse
MonkeySign
Assistance
PaddyPlantPrognosis
Progno D
Wonder
Architect
Volunteer Hub
Fundraiser
Sink or Swim
Pool Party
Remote Trainer
Tetris Duels
Tank
Building and deploying LLM agent systems for internal and partner-facing workflows: agent design, tool use, RAG integration, and production guardrails.
Partnering with product and engineering teams to set the evaluation, quality, and cost bar each solution must clear before launch.
Shipped a production multi-agent system that automates denied-claim investigations, with 3 sub-agents covering 20 policy workflows.
Scaled it to 300–500 weekly users and up to 20K claims a week.
Led agent design for a major news publisher's AI agent, launched publicly in Nov 2025 over a 102-year archive in 13 languages.
Owned the LLM-as-judge eval framework that took QA-audit accuracy from 64% → 88% and helped close a seven-figure expansion.
Cut the cost of a full health-system validation run about 4× ($4.6K → ~$1.2K).
Joined as a senior AI intern in June 2023 and converted to full-time that August.
Built an LLM analytics and evaluation pipeline that grew coverage of 10,000+ weekly support chats from 1–2% to 100%.
Cut processing time for that pipeline from 8 hours to 1.
Added rubric-based retrieval scoring and 20 monitoring dashboards, lifting accuracy 90% → 95% and dropping human hand-offs 40% → 13%.
Built an emergency-response AI platform with Berkeley SkyDeck funding.
Solo-engineered the low-latency voice agent and telephony behind the dispatcher demo: intent routing, emotion signals, and structured handoff to responders.
Won the Grand Prize at the UC Berkeley AI Hackathon against 900+ participants.
Earned $50K in total investment across two awards: the $25K Berkeley SkyDeck grand prize and a separate $25K investment from AIC.
Founded and shipped UCSC's real-time bus tracker, which served 20,000+ users.
Placed Global Top 10 out of 2,000+ projects in the Google Solution Challenge.
Taught Python through a challenge-based curriculum.
Coached USACO students from Bronze toward Silver through algorithm practice, debugging, and contest prep.
Competed mostly during college; still enter a few events a year.
Most of the projects on this page started as weekend builds.
Judged LA Hacks 2026.
Upcoming: confirmed as a judge for the LA Hacks AI Hackathon.
Graduate-level ML, applied AI, agentic systems. Built voice-first products throughout.
Finished a 4-year program in 2.5 years. Caught the hackathon bug.
South Bay. Arrived already coding, thanks to an iD Tech camp in middle school.
I build voice-first agents. I care about latency, keeping a human in the loop, and what happens when the model is wrong.
AI engineer: voice-first agents, multimodal pipelines, real-time inference.
Pinterest: LLM agent systems, RAG, production guardrails. Previously Scale AI.
USC: M.S. Computer Science, AI specialization.
Stack: TypeScript / Python / Rust, Next.js, Retell, Twilio, Mistral, Llama.
Ask about a project and the agent looks it up, describes it, and opens it on the page. Retell carries the audio, Deepgram hears you, GPT-5.6 Terra thinks, ElevenLabs talks back. A Python backend runs the agent and tells the browser where to fly.
Finished a 4-year CS degree in 2.5 years.
Won my first hackathon at 17.
Speaks Mandarin and English; building agents in both.
Plays drums and piano when the laptop is closed.
Still chasing the best bowl of noodles in San Francisco.
Convinced everyone gets a personal Jarvis eventually: an AI you just talk to, in real time. Building toward it.
Would rather talk to software than tap through it. Hence this site.
