Syed Tasavour — AI Engineer and Full Stack Developer

CH.01 — LIVERemote · ISTBudgam, Jammu & Kashmir, India

AI Engineer and Full Stack Developer. SaaS and web apps.

I take an unclear problem, work out what should actually get built, and then build the whole of it. Interface, API, database, deployment, and the alert that wakes me when it breaks.

Syed Tasavour — shipping production systems since 2019. 7 years.

Users served: 10M+; added up across everything shipped since 2019
Monthly actives: 3M+; Kashmir Student Alerts, a client platform. Not mine
Peak concurrent: 20K+; highest load held in production, on result days
Uptime: 99.9%; across the platforms I built
APIs shipped: 1,260+; REST endpoints in production since 2019

Building since 20197y 7m 9d66,686h4,001,189min240,071,385s240,071,385,560ms

01 — Trace

Four roles, one continuous recording.

01 / Jun 2025 — PresentBangalore, India

Beneathatree

Full Stack Developer

Speak2 connects people in senior living communities with their families, the staff and their caregivers. The same message has to reach an app, a website, an Alexa unit in the room and the TV in the lobby. That audience changes what counts as a bug: a missed notification is not a defect in a dashboard, it is a family that did not hear something. Also on the analytics side, where I moved the warehouse layer to a faster engine without breaking a single existing query.

Speak2 Family Uplift Beneathatree

Feature work across the stack on Next.js and Express: the communication flows, service requests and content delivery that residents and staff use daily
One set of APIs behind four surfaces (web, mobile, Alexa and the facility TVs) on Lambda, SES and VPC. Four clients off one contract, so a change lands everywhere or nowhere
Kept the signage pipeline feeding live content to the TVs in every building. It has no user to complain when it stalls, which is exactly why it needed monitoring more than the parts people can see
Added bounce monitoring and retry handling to the transactional email path on SES. Delivery failures are silent by nature, and a notification system people stop trusting gets ignored, at which point the feature may as well not exist
Debugging live production on a platform elderly residents and their families depend on, where a bad deploy is not an inconvenience, it is somebody not reaching their family
Serverless APIs on Lambda and API Gateway for the analytics side, plus the Next.js work in front of them. Event-driven and spiky, which is what made serverless the right shape rather than the fashionable one
Ran a warehouse migration onto a faster query engine, with the BI layer re-pointed on top of it. Speed was the reason it was funded; the requirement I actually cared about was that no existing query broke and nothing went offline. That is the only version of a migration a business will agree to

Next.jsNode.jsExpress.jsAWS LambdaAWS SESAPI GatewayVPCS3EC2Socket.IORedisPostgreSQLServerlessDocker

02 / Mar 2025 — Jun 2025Remote, India

Simmature

Full Stack Developer

Profiles, feeds, connections and messaging for a professional networking platform. Messaging was the part with a real constraint: a thousand people talking at once across more than one process means a socket on one instance has to reach a socket on another. Redis Pub/Sub between them was the simplest thing that worked, and simple mattered because encryption and voice calling sat on the same path.

Around 30 backend endpoints on Node, MongoDB and Redis covering profiles, feeds, connections and interactions
Messaging on Socket.IO with Redis Pub/Sub between instances, holding 1,000+ concurrent with end-to-end encryption and voice calling on the same path
JWT with Google and LinkedIn sign-in and 2FA on top; exponential backoff on Twilio and SendGrid to 99% delivery, a verification code that arrives late is worse than useless. The person has already tried again

Node.jsMongoDBRedisJWTSocket.IOAWS S3TwilioSendGridOAuth 2.0

03 / Oct 2024 — Nov 2024Remote, India

Unified Mentor

Full Stack Developer Intern

An education platform was showing every student the same static list of recommendations. I replaced that logic with a model reading the individual profile, so the suggestions changed per person and moved as the profile did. Also took about a quarter off the frontend load, and built the chat.

Replaced a static recommendation list with a model reading the individual profile, so what a student saw changed per person and moved as their profile did
Took about 25% off the React load through lazy loading, code splitting and memoisation; built the chat, which held 500+ concurrent

React.jsOpenAI APINode.jsMongoDBWebSocketsTailwind CSS

04 / Sep 2024 — Oct 2024Bangalore, India

Exposys

Full Stack Developer Intern

A travel marketplace with two thousand listings, deployed on EC2 behind auto-scaling. It was an internship and I am not going to inflate it. It was the first time I owned the deployment as well as the code, and finding out what auto-scaling does not do for you was more instructive than the application.

A travel marketplace with 2,000+ listings on Node, Express and MongoDB, with MapBox, accounts and reviews, on EC2 across five auto-scaled instances at 99.9% uptime

React.jsNode.jsExpress.jsMongoDBAWS EC2MapBox

02 — Manifest7 / 27 systems

Everything shipped, ranked by what it had to survive.

Most of these are live, in the Play Store, or on GitHub where you can read the code. A few are builds from 2024 that existed to get one specific thing right, and those say so.

03 — Signal

Most briefs describe a symptom. Somebody needs an app that does this, or their team loses every Monday to a spreadsheet. The first useful thing I do is find out what is actually wrong, which is often next to what was asked for. Sometimes half the scope is not worth the money, and saying so early is cheaper than saying it in month three.

Then I build it. All of it. Mostly serverless now, not because it is fashionable but because most of what I build is spiky and event driven. Kashmir Student Alerts sits quiet for days and then takes the whole region inside two minutes. That load pattern chose the architecture. It was not a preference.

Education: Bachelor of Computer Science, University of Kashmir
Working on: Multi-tenant SaaS · Event-driven backends · Real-time systems
Not a fit: Design-only work, or a spec that has already been decided
Status: Open to full-time roles, taking selective contract work

04 — Roll24 names · 2019—2026

Built for people who noticed when it broke.

Commissioned — 8

Shipped — 16

Hover to pause · hold and drag either way · arrow keys

05 — Domains

Systems for places where downtime is somebody’s day.

Ten domains since 2019. Each one below names the system behind it, so the claim and the evidence sit in the same line.

Senior living — Speak2 — voice, app, web and the TVs in each facility

Senior living
Sports analytics
Education
Non-profit
Fintech
Developer tools
Logistics
Publishing
Retail
Social

06 — Field53 tools

The tools I actually reach for.

Listed once so a screener can find what they are looking for. Everywhere else on this site a technology is named only when it explains why something was built the way it was. If a sentence about a project stays true after swapping Postgres for MySQL, the choice did not matter and it should not have been in the sentence.

Throw any chip

JavaScript — Languages, depth 95 of 100
React.js — Frontend, depth 95 of 100
HTML5 — Frontend, depth 95 of 100
Node.js — Backend, depth 95 of 100
Next.js — Frontend, depth 92 of 100
SQL — Languages, depth 90 of 100
Tailwind CSS — Frontend, depth 90 of 100
CSS3 — Frontend, depth 90 of 100
Responsive Design — Frontend, depth 90 of 100
Express.js — Backend, depth 90 of 100
REST APIs — Backend, depth 90 of 100
MongoDB — Database, depth 90 of 100
Socket.IO — Real-time, depth 90 of 100
Real-time Chat — Real-time, depth 90 of 100
Multi-tenant SaaS — Architecture, depth 88 of 100
Razorpay — Integrations, depth 88 of 100
AWS Compute (Lambda, EC2, API Gateway) — Cloud & DevOps, depth 86 of 100
Python — Languages, depth 85 of 100
Redux Toolkit — Frontend, depth 85 of 100
JWT — Backend, depth 85 of 100
API Design — Backend, depth 85 of 100
Redis — Database, depth 85 of 100
Database Design — Database, depth 85 of 100
Serverless Architecture — Cloud & DevOps, depth 85 of 100
WebRTC — Real-time, depth 85 of 100
Redis Pub/Sub — Real-time, depth 85 of 100
WebSockets — Real-time, depth 85 of 100
Webhooks — Integrations, depth 85 of 100
Event-driven Architecture — Architecture, depth 84 of 100
AWS Storage & CDN (S3, CloudFront, DynamoDB) — Cloud & DevOps, depth 82 of 100
DigitalOcean · Linux · Nginx — Cloud & DevOps, depth 82 of 100
Stripe — Integrations, depth 82 of 100
Java — Languages, depth 80 of 100
React Native — Frontend, depth 80 of 100
MySQL — Database, depth 80 of 100
Query Optimization — Database, depth 80 of 100
AWS Networking & Auth (VPC, Cognito, SES) — Cloud & DevOps, depth 80 of 100
Cloudflare (Workers, R2, D1, KV, Queues) — Cloud & DevOps, depth 80 of 100
Video Calling — Real-time, depth 80 of 100
Microservices — Architecture, depth 80 of 100
Twilio — Integrations, depth 80 of 100
SendGrid — Integrations, depth 80 of 100
Cloudinary — Integrations, depth 80 of 100
Mapbox — Integrations, depth 80 of 100
C — Languages, depth 75 of 100
Docker & CI/CD — Cloud & DevOps, depth 72 of 100
Supabase — Database, depth 70 of 100
NeonDB — Database, depth 68 of 100
Drizzle ORM — Database, depth 68 of 100
Fastify — Backend, depth 45 of 100
PostgreSQL — Database, depth 38 of 100
TypeScript — Languages, depth 35 of 100
Remix — Frontend, depth 35 of 100

07 — Ledger22 entries · all verifiable

Every credential, with the link that proves it.

Hover to pause · hold and drag · click any plate to read it full size

Education

Bachelor of Computer Science

University of Kashmir · Jammu & Kashmir, India · Completed

08 — JournalAll 8 entries ↗

Notes from things that broke.

Hover to hold · drag either way · click to read

09 — ContactSame day, weekdays

Still transmitting.

Tell me what is happening now, not what you have decided to build. What takes too long, what breaks, what somebody redoes by hand every week. That is the useful conversation, and it usually changes the scope. If it is a fit you get a scope with a number in it and something deliberately left out. If it is not, you get a reply saying so.

01 / Jun 2025 — PresentBangalore, India