I'm Joseph Zhang, a senior full-stack and AI engineer with 8+ years across React, Node.js, Python, Swift and cloud infrastructure. I built CodeMind Jobs on my own and have published 30+ apps on the App Store.
LLM pipelines, RAG, agents, speech and generative media, shipped in CodeMind, in 18 AI apps on the App Store, and at telecom scale at First Orion. Each one has real users and a cost budget.
CodeMind turns about 5,000 raw job posts a day into clean, typed data: skills, seniority, salary, visa and work type. It runs a cheap batch model first and moves up to a stronger model only when needed, and every answer is checked against a strict schema.
Chat with a PDF and get answers that cite the page, plus translation that keeps a book or paper's layout and formulas intact.
A research agent that plans, searches, reads and writes a report; an MCP server that lets AI assistants search CodeMind; and a browser agent that fills multi-page application forms.
Whisper transcription, natural text-to-speech with dozens of voices, and generated focus music, shipped as consumer apps with thousands of ratings.
In the ATS matcher the model only judges each requirement and quotes the resume line that proves it. Code checks that each quote is really in the resume and computes the score, and changes are tested against a benchmark of real job and resume pairs.
Photo-to-video animation, image editing by text prompt, photo restoration, virtual staging, background design and AI font art. Each one is a shipped App Store product with in-app purchases.
At First Orion: transformer embeddings and semantic matching for entity normalization, noisy-data matching and intent classification, feeding real-time scoring for caller identity and spam detection.
CodeMind gathers software-engineering jobs from hundreds of sources, scores your resume against each one, tailors it to the posting and fills the application form for you. I designed, built and operate the whole thing: ingestion, AI pipelines, web app, browser extension, billing and SEO.
Collectors for 34 ATS platforms (Greenhouse, Lever, Workday, Ashby, SuccessFactors…), plus LinkedIn, Indeed, hiring.cafe, Wellfound and YC. They run on schedules, merge multi-city reposts into one job and follow aggregator copies back to the employer's own link.
Every posting is turned into structured data (skills, seniority, salary, visa, work type) with a gpt-5-nano batch, then a retry, then gpt-5-mini, and every answer is validated. On flex processing it costs about $0.0005 per job.
The model only gives a verdict on each requirement and quotes the resume line that supports it. Code checks that each quote really appears in the resume, applies fixed weights and computes the score, so the result can be repeated and checked.
Strict-schema rewrites in three modes (ATS-first, Balanced, Looks-real), traced bullets, a gap pass driven by the score, and a benchmark of real job/resume pairs. Exports to ATS-safe .docx and PDF templates.
A Manifest V3 extension built from adapters for each board. It fills application forms that span several pages on Workday, Greenhouse, Lever, iCIMS and others, attaches the tailored resume and drafts answers, and leaves Submit to you.
Stripe billing, an employer job-posting flow, identity verification, job alerts via Resend, an MCP server for AI assistants, programmatic SEO landing pages and an admin console with outreach tooling.
Since 2015 I have designed, built, launched and monetized my own iOS and macOS apps, from language learning to on-device speech and generative AI tools, with 8,091+ ratings so far.
Chat with any PDF: retrieval-augmented answers with page citations, powered by GPT.
Turns any text, document or web page into natural speech with dozens of voices.
Generative AI that turns still photos into short animated videos.
Make, study and share flashcards with spaced repetition, images and audio.
On-device and cloud Whisper transcription for recordings, audio and video files.
A fast lookup-first vocabulary app with tags, fuzzy search and spaced-repetition reminders. My most-rated app.
I join teams as a senior engineer and take on select contract work. Three kinds of projects I do best:
RAG over your documents, agents that use your tools, structured extraction, and scoring you can audit. I build them with evals and a cost budget from day one.
From idea to a product people pay for: web app, backend, auth, billing, admin and deployment. The same path I took with CodeMind.
Native Swift/SwiftUI or React Native, designed for the App Store, with in-app purchases and on-device or cloud AI. I have launched 30+ apps of my own.
From the model call to the App Store release: the tools I reach for most.
I'm looking for senior full-stack and AI engineering roles. The fastest way to reach me is email.
