ABDUL REHMAN - SYSTEMS AND AI BUILDER

I build AI systems
that ship.

Agent harnesses, inference infrastructure, and automation pipelines. From a cookie-pool LLM backend to a 57-tool browser agent, built end to end and run in production.

Currently running: spidey.rehmthegreat.me inference backend, 7 shipped projects, 12 brands served.

01 / SELECTED WORK

Seven systems,
ranked by impact.

Every project below is live or shipped; the links go to the real thing.

Spidey Tracker - playable Spider-Man sightings tracker with CRT boot sequence and generated Manhattan city map

01 / SPIDEY TRACKER

Spidey Tracker

Playable Spider-Man sightings tracker, rebuilt in zero-dependency vanilla JS.

CRT power-on boot sequence. A full SVG lower-Manhattan-style city generated from data: 22 avenues, 34 cross streets, bridges, districts. Live WebAudio synth soundtrack. Drag, pan, zoom camera. Pure vanilla JS, zero dependencies.

  • Vanilla JS
  • SVG
  • WebAudio
  • Zero deps

02 / DEEPSEEK++

DeepSeek++

Harness-grade agentic Chrome MV3 extension over chat.deepseek.com.

WXT plus TypeScript. Intercepts completion requests, injects a 57 to 60 tool catalog into a site lacking native tool-calling, executes tool calls, inline agent loop with structured console, synchronous subagent spawns with nested consoles, persisted run traces.

  • WXT
  • TypeScript
  • Chrome MV3
  • Agent harness

03 / REHM'S INFERENCE + AGENT HARNESS

Rehm's Inference + Agent Harness

Self-hosted LLM inference backend serving spidey.rehmthegreat.me.

Cookie-pool front (ClewdR), OpenAI-compatible endpoint, independent VPS and PC MCP gateways with lazy tool discovery, agentic harness on top with a 496-check selfcheck suite. Runs in production today.

  • ClewdR
  • MCP gateways
  • OpenAI-compatible
  • 496 checks
496 selfcheck tests passing

04 / DOCTORMARKETING

DoctorMarketing

Social Media Creative Engine. Raw brand assets in, published 1024 by 1536 creatives out.

Turns raw brand assets into standardized brand.md profiles and published social creatives. Web scraper for OpenGraph and DOM colors and fonts, strategic content calendar generation, Colab T4 image generation, n8n orchestration, web UI. Runs as a private pipeline.

  • n8n
  • Colab T4
  • Web scraper
  • Brand profiles

05 / CONTENT ENGINE / REHMARKETING

Content Engine

The agency track record, presented as proof.

12brands served
147pieces per week
4accounts owned
3people led

Human-in-the-loop gates, n8n pipelines, multi-brand content calendars, batch image generation. Ran social for 12 brands at Rise High Tech, owned 4 client accounts, led 3 people, throughput 147 content pieces per week across 12 brands. First independent client landed via own pipeline (Asian Continental).

06 / FOODPANDA-AI-MCP

Foodpanda AI MCP

Model Context Protocol server for foodpanda, public repo with CI.

An MCP server exposing foodpanda operations to AI agents. Public repository with continuous integration. Built so an agent can search, order, and track on behalf of a user through a standard protocol.

  • MCP
  • Node.js
  • CI
  • Public repo
MCP protocol server

07 / PSX MARKET WATCH SIGNALS

PSX Market Watch Signals

Automated Pakistan Stock Exchange signal pipeline.

Live market data, portfolio analysis, rule-based signals, AI-formatted email alerts. n8n plus OpenAI plus Google Sheets plus Gmail. Runs on a schedule and emails a digest without manual steps.

  • n8n
  • OpenAI
  • Google Sheets
  • Gmail

02 / PROOF

Systems that run
in production.

Measured from shipped work.

6 years SRE and ML-infra at a hyperscale cloud
BSc Computer Science
12 brands run at Rise High Tech
147 content pieces per week throughput
496 selfcheck tests on the inference harness
7 shipped projects

03 / ABOUT

Close to the problem.
Hands on the build.

I'm Abdul Rehman, Rehm the Great. I build AI systems, automations, and backend tools that solve real operational bottlenecks.

My focus is practical engineering over hype: deterministic state machines where reliability matters, generative AI where unstructured synthesis helps, and clean glue between systems that refuse to talk to each other.

I work iteratively. Prototype the core loop, test it against real workloads, keep human oversight where judgment counts. Everything I build is transparent, observable, and dependable.

A /

AI and LLM systems

Agent harnesses, tool catalogs, inference backends, cookie-pool fronts, MCP gateways, context engineering.

B /

Automation

n8n pipelines, webhook flows, content engines, scrapers, scheduled digests, human-in-the-loop gates.

C /

Backend and tooling

Node.js, Python, TypeScript, Chrome extensions, MCP servers, Linux systems, self-hosted infra.

Start a conversation

04 / CONTACT

Have a system to build?
Let's talk.

rehm@rehmthegreat.me

Open for AI system builds, agent harness work, inference infra, and automation consulting.