Sayyed Arib
Hussain_

Software engineer. I'd rather read the config than guess at it.

A year and a half building and operating a CRM that 100,000+ people use every day — Angular, Node.js, PostgreSQL. I contribute upstream to open source, automate the parts of my job that repeat, and care more about why something broke than about closing the ticket.

Experience

~/experience

Software Engineer, telecrm

Jan 2025 — Present

Full-stack CRM platform · Gurugram, India (onsite)

7,000+ clients · 100,000+ users · 745M+ API requests a day

  • Rebuilt role-based access control across 5 modules — dashboards, message templates, saved filters, embedded apps, and confidential fields — so permissions decide what every role can see and edit. Shipped two-factor authentication alongside it.
  • Wrote 300+ automated regression tests (Puppeteer, Jest) that run in CI on Ubuntu runners and gate every release, including a reusable helper for driving elements outside the viewport that the team still uses.
  • Built a Model Context Protocol integration with 6 read/write tools, letting LLM applications drive internal CRM workflows.
  • Sat with stakeholders to work out what they actually needed, then delivered 200+ analytics dashboards for 40+ clients. One funnel view showed a marketing team where users dropped off and helped lift campaign reply rates 20%.
  • Own the ETL pipeline that moves data from MongoDB into Amazon RDS for analytics — scripts scheduled with systemd timers, feeding a self-hosted Metabase instance on Docker.
  • Take first-line support duty every fourth sprint: triaged 50+ tickets, fixed what I could, routed the rest, and coordinated with product.
  • Own production incidents end to end — reproduce, trace the root cause across the stack, ship the fix.

Full-Stack Developer Intern, VoxxiLab Inc.

Apr — Jul 2024

AI voice-monitoring product · Remote

  • Built backend APIs in Python, FastAPI, and PostgreSQL behind lead-management and voice-monitoring dashboards (React, Supabase).
  • Tuned LLM workflows for sentiment analysis and call-quality review, cutting pipeline latency by ~200 ms.

Open source

~/open-source

zulip / zulip

5 pull requests merged into the production codebase. The one I'm proudest of is a bot-enforced access-control feature that stops people claiming a pull request they aren't assigned to. I also added silent mentions to cut notification noise and improved file-upload UX with a drag-and-drop effect.

hacktoberfest

I started contributing while preparing for Google Summer of Code and kept going. Contributions across 8+ organizations, working through code review in large codebases I'd never seen before — including one merged bug fix to a Rust project. A small one, but my first.

  • Layer5 / Meshery
  • Ockam
  • Appwrite
  • Monday.com
  • AMUpedia

why I stayed

I came for the programs and stayed for the people. In Zulip and Meshery, founders and senior engineers took time to point out my careless mistakes politely — that's what pushed me to get better. I still read through projects I'm not contributing to, just to see how others handle scale and maintenance.

Projects

~/projects

Khan Group of Hostels & PG

still in production
Next.js · Supabase · Drizzle

A booking platform for a hostel owner I met while staying there during college — my first large solo build. 150+ bookings and about INR 75,000 in its first two weeks. The lesson I actually took away was about requirements: scope kept moving with the client's mood until I stopped and froze it, sketched the UI on paper, then designed the database.

The hard part was double-booking. Two people could hold the same room while a payment sat unverified, so I used a database transaction to make sure only one booking can ever lock a room, even under concurrent requests.

Grabtern

early core contributor, not the creator
Next.js · MongoDB · Node.js

A senior of mine started this mentor–mentee platform and I joined a few weeks in, during my fifth semester. I built the mentor registration page and Google login, then took over more of the core engineering — starting with the Razorpay payment gateway — as he moved to the business side. We onboarded 50+ mentors and 20+ mentees, and the community page still has 6,300+ followers on LinkedIn.

It was picked for GirlScript Summer of Code, which brought in a wave of first-time contributors. I spent much of that time reviewing their PRs and walking them through small mistakes — the same patience Zulip's maintainers had shown me. The project is shut down now; Topmate took that market.

Botcraft

Next.js 15 · React 19 · FastAPI · LangChain

A no-code platform for building AI assistants: upload a knowledge base, tune the retrieval pipeline, test it live, then paste one <script> tag into your site. I designed the LLM orchestration layer and the backend, and built the dashboard around it — 22 routes covering auth, onboarding, a multi-source knowledge base, analytics, a WebSocket playground, and a theming studio.

The part I'd defend in a review is the state boundary. Server state lives in TanStack Query with explicit invalidation; client-only state lives in a persisted Zustand store. Neither leaks into the other, so no component ever calls Axios directly — one hook per resource owns its query keys.

The whole RAG pipeline is configurable from the UI at runtime: chunk size, overlap, splitter, PDF parser, embedding model, LoRA/QLoRA parameters, temperature, and the system prompt — every field validated with Zod, which also infers the TypeScript types. Auth rides on an HTTP-only cookie, so the token is never readable from JavaScript. CI runs lint, tsc --noEmit, Vitest, and a production build on every push; the tests guard the things that actually bite, like an https:// page deriving wss:// instead of a mixed-content ws://.

Project Alpha

Vercera Hackathon 2024, AMU
Python · TensorFlow · MoveNet · Next.js

A way to play computer games with your body instead of a keyboard. MoveNet reads your pose from the webcam and maps it to game input — raise your left hand, the character turns left — so time at the machine costs you some movement instead of none. Pose detection in TensorFlow and OpenCV, input synthesis with PyAutoGUI, Next.js front end.

Home automation by speech recognition

Python · Machine Learning · Arduino

An end-to-end pipeline rather than a demo: collected and labelled a 1,000-sample voice dataset, trained and evaluated a speech-command classifier, then deployed it to an Arduino for real-time inference — switching lights on a spoken command.

Skills

~/skills
languages
JavaScript, TypeScript, SQL, C++, Java · Python (FastAPI, pandas — rusty but real) · Go, Rust (exposure)
frontend
Angular, React, Next.js, HTML, CSS, reusable components
backend
Node.js, FastAPI, REST APIs, PostgreSQL, MongoDB, Redis
linux & infra
Ubuntu, Bash, systemd services and timers, networking / DNS, file permissions, Docker, AWS (RDS)
testing
Puppeteer, Jest, regression automation, root-cause debugging, incident triage
data & ai
ETL pipelines, Metabase, query optimization, LangChain, Model Context Protocol, RAG, NumPy, pandas
practices
Git / GitHub, GitHub Actions, code review, requirements gathering, on-call rotations, documentation

Education

~/education

B.Tech, Artificial Intelligence — Aligarh Muslim University

CGPA 8.4 / 10

Dec 2021 — Jun 2025

Class XII, CBSE — Aims International, Gonda, UP

94.2%

  • Crash Course on Python — Google / Coursera
  • Supervised Machine Learning: Regression and Classification — Stanford / DeepLearning.AI

Get in touch

~/contact

Open to backend, full-stack, and platform roles. If you've read this far, the fastest way to reach me is email.

sayyedaribhussain4321@gmail.com
arib@ubuntu:~$ ↑ history · tab completes