HashirHussain
Digital portrait of Hashir Hussain

The restless builder

Software engineer
AI consultant.

I build products end-to-end and make the software your business already runs more intelligent.

Explore the systems

I build systems that

Think.

Live workflow / 01

The system
behind the button.

A customer asks one question. The system gathers context, reasons inside guardrails, takes action and knows when a person should step in.

WhatsApp · CRM · retrieval · approval · Power Automate

Customer support / production pattern

01Event

A message arrives

  • Where is order #4821?
02Context

The record assembles

  • Customer
  • Order #4821
  • Delivery policy
03Reason

Understand. Retrieve. Decide.

  • Understand
  • Retrieve
  • Decide
04Guardrail

Confidence before action

  • Confidence check
  • Human approval

Exception pathHuman review

05Action

The tools do the work

  • Reply
  • Update CRM
  • Trigger flow
06Result

The routine work disappears

60–70%of customer queries resolved without a human, in production.

Proof / 02

Proof,
not promises.

A system earns its place when it answers faster, costs less, or scales further.

0160–80%

Infrastructure cost reduction

Quinn · AI commerce SaaS

02<15s

First response to customers

Quinn · AI commerce SaaS

03~60–70%

Queries resolved without a human

BOOST Lifestyle · AI CRM

0410,000+

Dynamically generated expert agents

Apna Tutor · Mustaqbil Hackathon 2026 winner

Work / 03

Systems
in production.

Products I built and AI I embedded into software people already use.

  • 01

    MCP connector · Own product · Live beta

    ReelVision

    Your AI can finally watch reels. Paste an Instagram, TikTok, Facebook or X link into Claude and ReelVision hands it the keyframes, the caption and a timestamped transcript, so a saved reel turns into notes, steps or a build plan.

    Live: 1,006 reels watched, 124 signups since 9 Oct

    • Free beta · works in Claude
    • Keyframes + timestamped transcript
    • Runs on one $6 VPS
    • FastMCP · Groq · yt-dlp · PostgreSQL · Google OAuth

    What Claude receives

    01 / Link

    Opening frame of the ReelVision ad: a cartoon Hashir sitting on a pile of saved reels.linkedin.com/posts/hashirhussain-ai…0:52.8

    02 / Keyframes

    1. A pile of 1,000 saved reels0:02.0
    2. An AI chat replying “Sorry, I can't watch videos.”0:22.0
    3. The reel splitting into keyframes0:28.0
    4. A question: “How do I build this?”0:36.5
    5. “Building it yourself… 3% · eta: 6 months”0:44.0
    6. ReelVision card: “Your AI can finally watch reels.”0:49.8

    03 / Transcript

    1. [0:00.1] If you saved 1,000 design reels, would you become a better designer?
    2. [0:18.9] You could paste the link into an AI, but it can't watch videos. It just sees a link.
    3. [0:23.6] Pass that reel through a special connector and it gets split into frames. Every word becomes text.

    04 / Answer

    How do I build this?

    It's a 53-second explainer: a saved-reels problem, an assistant that only sees a link, then a connector that turns the reel into keyframes and a transcript. To build the same pipeline, fetch the public video, sample keyframes, transcribe the audio with timestamps, and hand all three to the assistant over MCP.

    Illustrative answer

    Example: ReelVision's own launch ad, run through the connector.
  • 02

    AI SaaS · Multi-restaurant ordering

    EasyOrder

    One chat, many restaurants. Customers order from any restaurant on the platform in a single conversation, with nothing to install and no login. An AI agent takes the order in Roman Urdu or English, and each restaurant runs its orders from a live dashboard.

    • Many restaurants, one conversation
    • AI ordering agent, Roman Urdu + English
    • Restaurant and admin dashboards, realtime
    • OpenAI Agents SDK · FastAPI · Next.js · Supabase
  • 03

    AI SaaS · Commerce operations

    Quinn

    One operating system for an e-commerce business, run by two AIs. A support AI answers customer queries and takes orders, and an ops AI runs operations alongside the team. Both use retrieval over the business's own data and understand text and images.

    • 60–80% infrastructure cost reduction
    • <15s first response to customers
    • Shopify, WooCommerce, Leopards, PostEx, Stripe
    • Multi-tenant · RAG · Multimodal · FastAPI

7 more builds: MCP servers, voice agents, hackathon wins and client systems.

See all work

Experience / 04

Where I
build.

  1. Jun 2024 — now

    Forward Deployed Engineer

    Freelance

    I embed with a client's team, learn how the business actually runs, and ship the AI system into it, from first call to production.

    • EasyOrder Multi-restaurant AI ordering SaaS: one chat, many restaurants
    • Outreach OS Autonomous lead research and personal emails, with a human approving each send
    • Corporate task assigner Assigns tasks and sends notifications across a whole company, from inside Claude, through a remote MCP connector
    • PISM Inventory and production system for a lubricant manufacturer
    • HRM Agentic HR system with five agents, with a human approving every outward action
    • Canadian venture group One parent site that presents child care, senior care, innovation and IT ventures as a single institution
    • US corporate Strategic data analysis and insights review across modern CRMs: Salesforce and HubSpot
  2. Oct 2025 — Jul 2026

    AI Full-Stack Engineer

    Extensor Labs · a Marktecs company

    Built the company's AI products end to end.

    • Quinn Multi-tenant AI commerce SaaS
    • Water-management systems company Bilingual voice agent that takes orders and opens complaint tickets
    • Digitemb AI lead follow-up that closes orders over email for an embroidery-digitizing and vector-artwork company serving international clients
  3. Oct 2025 — Jul 2026

    AI Full-Stack Engineer

    Marktecs

    Led AI and backend.

    • BOOST Lifestyle AI CRM with broadcasting, a referral program and an agent that resolves ~60–70% of customer queries on its own
  4. Student

    BS Computer Science

    University of Karachi

    Studying while shipping.

Skills / 05

The
toolkit.

Languages

  • Python
  • TypeScript
  • JavaScript

Web & backend

  • Next.js
  • React
  • FastAPI
  • PostgreSQL
  • Docker

AI

  • OpenAI
  • Gemini
  • RAG
  • Agents
  • MCP

Agentic coding

  • Claude Code
  • Codex

Commerce & messaging

  • Shopify
  • WooCommerce
  • WhatsApp
  • Stripe

Business platforms

  • Salesforce
  • Dynamics 365
  • HubSpot
  • Power Automate
  • SharePoint

LinkedIn / 06

Building
in public.

What I'm shipping and learning, as it happens.

Human / field notes

Built with
people.

The work starts by making the workflow visible—on a whiteboard, around a table, and with the people who run it.

Think and explain

Before code, make the workflow visible. Then make the trade-offs clear to the room.

Hashir mapping a workflow beside a whiteboard and display
Think

Map the workflow.

Hashir presenting a software system to colleagues around a table
Explain

Explain the system.

Build and practise

Production work gets better when code, operations, and feedback meet at the same table.

Hashir and his team reviewing code together on a laptop
Build

Build together.

Hashir holding a team certificate outside the National Aerospace Science and Technology Park
Practise

Keep showing the work.

Prove and explore

Test ideas under pressure, show the result, then keep looking for the next useful problem.

Hashir and teammates with their OneAI hackathon winner cheque
Prove

Ship under pressure.

OneAI hackathon winner

Hashir at the Multi Domain AI and Innovation Summit
Explore

Stay curious.

Multi Domain AI & Innovation Summit

Have a workflow that keeps stealing human time?

Let’s redesign it.

We’ll map the work, find the decisions, and build the system behind it.

Start a project