about

Mohammad Shadiul Huda

I'm Shadiul Huda — Utso to most people.

I'm a Senior Software Engineer and the frontend lead at Mevrik, working from Dhaka, Bangladesh. We build AI agents that run customer experience while humans supervise what matters — I own the frontend and the range of surfaces it takes: the operator console and its omni-channel inbox, a CRM and segment builder, an embeddable chat widget with a one-line install script, an n8n-style workflow canvas, an AI agent studio, and the analytics dashboards — all on the shared UI kit they're built from. I've led the frontend for 2+ years and mentor the team through code review.

It goes past the UI. I own how the frontend ships — Cloudflare, CI/CD and Docker, handed clean to devops — and I build the platform's voice and video AI: an AI-driven call center on Voximplant (voicebots, campaigns, transcription, call summaries) and in-app video with screen share and device control. Plus the internal devtools the team runs on.

Nights are for systems. The one I keep coming back to is my second brain — a knowledge base Claude operates for me, where new information rewrites old pages and scheduled agents reconcile the vault while I sleep. This site is another one of those builds: it doubles as the build log.

I work across models and providers — Claude, GPT, Gemini, DeepSeek, and local Llama or Gemma on Ollama — and the craft I care about most is context engineering: the right information in front of a model at the right moment, with MCP, agent skills and workflows doing the wiring.

I care about interfaces that stay fast, systems that stay honest, and notes that stay current without me babysitting them.

Path here

  1. 2022 — present
    Senior Software Engineer & frontend lead, Mevrik — Dhaka
  2. 2021 — 2022
    Software Engineer, NexusLab — fintech — real-time trading dashboards, web + mobile
  3. 2020 — 2021
    Mobile App Developer, Antooba Corp
  4. 2018 — 2019
    Unity / AR-VR Developer, GrapView
  5. 2017 — 2021
    B.Sc. Computer Science & Engineering — Daffodil International University

Where I'm headed

Lately I'm pulled less toward shipping features and more toward the engineering principles that make software trustworthy — software verification, formal methods, and metamorphic testing for the AI/ML parts that are hardest to pin down.

My angle on it is human. Navigating an English-first industry as a native Bengali speaker taught me how much interface and language decide who technology actually works for — so I care about this most for multilingual and accessibility-critical systems. Long term, I want to build HCI and NLP tools tailored to needs like that, back home.

software verificationformal methodsmetamorphic testingHCIaccessibilitymultilingual NLP

Publications

3 papers

What I work with

depth of use — core to occasional
Languages daily
JavaScript TypeScript Python
Frontend daily
React Next.js Astro Redux Zustand MDX
UI & styling daily
Tailwind shadcn/ui Base UI Radix
Data & forms daily
TanStack Zod React Hook Form Axios
Backend & data integrate
Node.js Hono PostgreSQL MongoDB
Realtime & voice build
WebSockets Voximplant WebRTC
Infra & build ship
Cloudflare DigitalOcean Docker CI/CD Turborepo Vite bun pnpm yarn Firebase Git
AI & agents daily
Claude GPT Gemini DeepSeek Llama Ollama OpenRouter MCP n8n
Tools daily
VS Code Postman Slack Obsidian

How I work

Verify live, not just green

Passing typecheck and CI is not "done". Done is clicking through the real thing against real APIs.

Don't fake it, don't delete it

If a feature is not ready, disable it honestly. Never ship a UI that pretends, and never rip out work that still has a job to do.

Ask before deciding

Ambiguous product calls get surfaced before the change lands — nobody should discover a decision after the fact.

Notes that write themselves

If knowledge only lives in my head, it is a liability. Everything worth keeping goes into a system that keeps it current.

Recognition

Say hello