01
AI agents & RAG systems
Production LLM pipelines with evaluation built in: retrieval, tool use, LLM-as-judge scoring and self-hosted open models when data cannot leave your network.
- Python
- FastAPI
- Gemini / GPT / Groq
- Qwen-VL
- PyTorch
Taking on select contract work
Software that ships,
and holds up
under audit.
I am Muhammad Hassan, Full Stack / AI Software Engineer. I build accessibility tooling, AI agents and RAG systems for US startups.
- 3 yrs
- Remote engineering for US companies
- 10K+
- App Store users on PackRat AI
- 100%
- Job Success Score on Upwork
- 508
- DHS Trusted Tester certified
What I do
01
Three things I am hired for.
Each one comes with the same habit: measurable outcomes, tests you can run, and documentation your team will actually read.
02
Accessibility engineering
WCAG 2.2, Section 508 and European Accessibility Act compliance that is actually verified: rule engines, automated audits and remediation, not checkbox reports.
- WCAG 2.2
- Testaro
- Playwright
- DevTools a11y tree
- APCA
03
Full stack web & mobile
Next.js and React Native apps shipped to the App Store and Play Store, with Node or .NET backends and the payments, messaging and maps integrations startups need.
- Next.js
- React Native
- Node.js
- tRPC
- .NET
- AWS
Case studies
02
Work, written the way a client reads it.
Problem first, then what I built, then what changed. No screenshots of dashboards you cannot log in to.
01 · 2025 — 26
AI-enhanced WCAG rule engine
YRA Tech · Lead engineer
Internal product, details on request
- Problem
- Testaro's accessibility rules flag issues but cannot say how severe they are or how to fix them safely.
- Built
- Reimplemented rules such as titledEl, nonTable, allCaps and language-of-parts as Python APIs where a model classifies severity, picks a fix from a closed vocabulary and explains itself, while deterministic patchers and a verification oracle handle auto-remediation. Shadow-mode runner plus a thin Node.js client over JSON stdio.
- Result
- Rules now ship with explainable severity and safe, verified patches instead of raw flags.
- Python
- BeautifulSoup
- Gemini Flash
- Node.js
- Playwright
- pytest
02 · 2025
Neural translation comparison tool
YRA Tech · Full stack / AI engineer
Internal product, details on request
- Problem
- Choosing a machine-translation engine per language pair was guesswork.
- Built
- Two-way comparison of DeepL, Google Translate, NLLB-200 and Opus-MT with MBR decoding, LLM-as-judge scoring via Groq, word-level LCS diffs, back-translation confidence and a terminology glossary.
- Result
- Side-by-side, scored output that makes engine selection a data decision.
- Next.js 14
- TypeScript
- Python
- NLLB-200
- Opus-MT
- DeepL API
- Groq
03 · 2025
Visual-semantic mismatch detection
YRA Tech · ML engineer
Internal product, details on request
- Problem
- Pages can look fine while their accessibility tree tells screen readers something else entirely.
- Built
- Paired Chrome DevTools accessibility trees with page screenshots and vision-language models to detect where the visual and semantic layers disagree.
- Result
- A new class of defects surfaced that no rule-based checker catches.
- Python
- Chrome DevTools Protocol
- Qwen-VL
- Playwright
04 · 2024
PackRat AI travel app
Bierman Collective · Full stack developer
- Problem
- A trip-planning product needed web, iOS and Android from one small team.
- Built
- Shared codebase across Next.js and React Native using Solito, tRPC and Drizzle ORM, Mapbox routing, Tamagui UI and Firebase auth.
- Result
- Live on both stores with 10K+ App Store users, operational across Europe.
05 · 2024 — 25
Devellp CRM platform
Devellp · Software engineer
- Problem
- A field-service startup needed a mobile CRM with payments, messaging and video built in.
- Built
- Owned the React Native app end to end alongside a .NET and Python microservices backend with Apple Pay, Twilio and Zeego integrations.
- Result
- Shipped and live on the US App Store.
- React Native
- TypeScript
- .NET
- Python
- MongoDB
- Apple Pay
- Twilio
06 · 2023
Real-time chat
Internee.pk · Intern
- Problem
- Internship brief: one-to-one chat with proper auth.
- Built
- Express and MongoDB API, JWT authentication, Socket.IO transport, Chakra UI front end.
- Result
- Delivered within the internship window.
- React
- Express
- Socket.IO
- MongoDB
- JWT
Experience
03
Three years, all remote, all US clients.
Sep 2025
— PresentFull Stack / AI Software Engineer · YRA Tech
Massachusetts, USA · Remote
- Reimplementing Testaro accessibility rules (titledEl, nonTable, allCaps, language of parts) as AI-enhanced Python APIs with deterministic patchers and a verification oracle, plus a Node.js client.
- Built a two-way neural translation comparison tool (DeepL, Google Translate, NLLB-200, Opus-MT) with MBR decoding, LLM-as-judge scoring and back-translation confidence, shipped as a Next.js 14 app.
- Scoping perceived-contrast-lab, a research apparatus measuring human perception of text contrast, aimed at a formula competitive with APCA for future WCAG standards.
- Built visual-semantic mismatch detection from Chrome DevTools accessibility trees and screenshots; contributed to a WCAG rules gap analysis against Testaro engines.
- Site admin on YRA Monitor, the company's continuous WCAG 2.2 / EAA scanning and remediation product.
Oct 2024
— Jul 2025Software Engineer · Devellp
USA · Remote
- Joined as intern, promoted to engineer in two months; owned the React Native app of a full CRM platform live on the US App Store.
- Worked across .NET, Python, Angular, MongoDB, Apple Pay, Twilio and a microservices backend.
Feb 2024
— Sep 2024React Native Developer (Full Stack) · Bierman Collective · PackRat AI
USA · Remote
- Designed the UI and built server routes for the PackRat AI travel app; shipped iOS and Android builds now past 10K App Store users, operational across Europe.
Sep 2023
— Oct 2023React JS Intern · Internee.pk
Pakistan
- Built a real-time chat app with Socket.IO, Express, MongoDB and JWT auth.
Aug 2021
— PresentFreelance Software Engineer · Upwork and direct clients
Remote
- 100% Job Success Score on Upwork as an AI Agent & RAG Engineer for full-stack SaaS.
- Hands-on with Retell, Vapi, Bland, Twilio and GoHighLevel for AI voice-agent projects.
Capabilities
04
The toolkit, grouped by the job it does.
Frontend by training, full stack by necessity, AI and accessibility by choice.
Frontend10
React
Next.js- React Native
Angular
TypeScript
Tailwind CSS
Material UI- Tamagui
- Redux
- Zustand
Backend10
Node.js- Express
- tRPC
Python- FastAPI
- .NET
GraphQL- Socket.IO
- Prisma
- Drizzle ORM
AI / ML9
- LLM pipelines
- RAG
- Multi-agent systems
- Vision-language models
- Gemini / GPT / Groq
- NMT (NLLB-200, Opus-MT)
- MBR decoding
- LLM-as-judge
PyTorch
Accessibility8
- WCAG 2.2
- Section 508
- European Accessibility Act
- Testaro
- ANDI
- DevTools a11y tree
Playwright- APCA research
Voice & automation5
- Retell
- Vapi
- Bland
- Twilio
- GoHighLevel
Data & cloud8
MongoDB
MySQL
PostgreSQL
Firebase
AWS
Docker- Tailscale
- Linux / WSL
About
05
Self-taught first, degree second, certified third.
Three years of remote work for US companies. Currently at YRA Tech, a Massachusetts accessibility company, building AI-enhanced WCAG rule engines, neural translation comparison tooling and contrast-perception research. Before that I shipped a CRM platform for Devellp and the PackRat AI travel app, now past 10K App Store users, for Bierman Collective.
I started writing code in 2021 and have not stopped. Outside client work I read accessibility standards for fun, solve LeetCode problems, answer questions on Stack Overflow and write about what I learn.
Research
Automated audio description for video · 2026
Ranked strategic recommendation for a training-free VLM-to-LLM pipeline, with a joint timing optimizer and on-screen-code triage as differentiators.
Perceived text contrast · 2026
Measuring human perception of text contrast to inform a next-generation contrast formula for WCAG.
Certifications
DHS Trusted Tester for Web (Section 508)
U.S. Department of Homeland Security · 2026
IELTS Academic, overall band 6.5
British Council / IDP
Education
2021 — 2025
BS Software Engineering
The Superior University, Lahore
2019 — 2021
Intermediate (HSSC), General Science
Government Post Graduate College, Jhelum
2017 — 2019
Matriculation (SSC), Science
Government High School, Jhelum
Writing
06
Notes I publish as I learn.
Hire me
07
Here is exactly what happens next.
Taking on select contract work. Tell me about the product and I will come back with a written plan, not a sales call.
- 01
Scope call
Thirty minutes. You explain the problem, I ask the awkward questions early.
- 02
Written plan
Milestones, risks and a fixed first deliverable within two days of the call.
- 03
Build in the open
Weekly demos on a staging URL, PRs you can read, no surprises at the end.
- 04
Hand-off
Docs, tests and a recorded walkthrough so your team owns it after I leave.
Prefer email? engineermirzahassan@gmail.com