local ai · sovereignty · never pay again

Asnan Abidi

I build AI that runs on your own hardware.

Local agents & servers, deployed on hardware you own.
Your data stays yours — and you never pay to use it again.

↓ scroll — the good stuff is below

You're renting AI. Own it instead.

The cloud wants you to pay every month, forever, and hand over your data in the bargain. Here's the other way.

the usual way ✗
  • API tokens£480/mo
  • more tokens£820/mo
  • your data, their serveralways
  • the pricewhenever they say
total∞ / forever
my way ✓
  • hardware, curatedone-time
  • server, builtone-time
  • your datanever leaves
  • usage£0 forever
total£0 — yours
Sovereignty Privacy Never pay again Decentralise AI

Proof, by the numbers

Every token, locally. Here's what that looks like — and what it would've cost in the cloud.

0
curated, locally deployed AI projects delivered
0
monthly users of finished solutions
0
tokens spent — £0 total cost

The cloud bill I never got

Same tokens, someone else's GPU:

Claude · Sonnet$9 / 1M
$0
28,456,098,357 tok × $9/1M
paid £0 ✌
ChatGPT · GPT-4o$6 / 1M
$0
28,456,098,357 tok × $6/1M
paid £0 ✌
Gemini · 2.0 Pro$5 / 1M
$0
28,456,098,357 tok × $5/1M
paid £0 ✌
Premium · Opus / 4.5$30 / 1M
$0
28,456,098,357 tok × $30/1M
paid £0 ✌
replay it at:
I paid £0 — all of it on my own hardware

From a gaming PC to a multi-agent system

One year, zero cloud bills. The short version.

Sep 2025
Arrived in the UK. 140 applications per vacancy. A degree wasn't going to be enough — so I started building.
Nov 2025
First local model in LM Studio. No idea what quantisation meant. Learned everything from Reddit.
Jan 2026
Early adopter of Open Claw. A local AI agent, running on my laptop. This was real.
Feb 2026
Unplugged the monitor. Pandora became a headless AI server, accessed only over SSH and Tailscale.
Mid 2026
Self-hosting everything: RAG bots, Docker, a mesh VPN, a multi-agent team with names and jobs.
Late 2026
First real client: AskJoe. Built entirely on local AI. Zero cloud costs.

Meet the team

Not one AI doing everything — a system of specialists, each with a job. Hover them.

chief orchestrator

Frankenstein

Delegates tasks across the whole system through Telegram group chats. The one in charge.

PandoraGemma 4 26B
the builder

OpenCode

Code review, debugging, refactoring — the full development lifecycle, inside Docker.

PandoraQwen 3.6 35B
the analyst

Helios

Deep reasoning, research synthesis, and decomposing hard problems.

PandoraQwen 3.6 27B
the planner

Super Sage

Runs Claude Code — the system's reasoning layer that plans the work and guides the other agents.

PandoraClaude Code · Docker
the pa

Agent Tim

Runs the MacBook: meetings, voice, transcription, computer use.

TimGemma 4 26B
the reader

Beacon

Reads the news and delivers your daily briefing.

TimGoogle Flash · Docker
hr & ops

HR Rick

Payroll, benefits and people data — the HR brain that keeps the admin moving.

TimQwen 3.6 27B

Proof, not promises

My first real client: AskJoe, a WhatsApp bot that answers questions from a professor's published work.

a colleague
"What's Joe's take on consulting?"
AskJoe · local
▌
Built on local AI. Zero cloud costs. Now being fine-tuned with a model I distilled myself.

The day job

Before and around the agents — payroll, people-ops and analytics. The part that taught me to spot what's broken.

Freelance AI Developer2025 – Present
Private client — leading UK business consultant & author
  • Production AI assistant, 500+ queries/month over WhatsApp & webchat, zero cloud inference costs.
  • Tool-calling RAG pipeline: 199 documents → 2,381 chunks, validated with 200–300 query eval runs.
Team Lead, Consultancy Project2026
Cardiff University MSc Consultancy Challenge
  • Led 5 people on a £1.8m charity analytics engagement — reconciled 429 survey responses against 1,395 bookings.
  • Flagged a VAT compliance risk and shipped a local AI agent that saved 50+ hours.
Core Team, HR & Operations2023 – 2025
Broseph Foundation (NGO)
  • Payroll inputs, benefits and audit-ready records across a 12,000+ volunteer network.
  • BambooHR data integrity: +15% reporting accuracy and monthly leadership dashboards.
HR Manager2023 – 2025
Body Science (healthcare)
  • Payroll & HR in Paycom/Workday for a 60+ person healthcare business.
  • Coached 4 managers, cut formal grievances 30%, 100% review completion twice running.
MSc Business Management — Cardiff University BSc Psychology — M. S. Ramaiah University

Hire me

From your payroll data to your own AI — I do the whole thing, on your hardware.

Swap a monthly bill for a machine that's yours.

Let's talk

Paying for AI every month — or drowning in payroll admin? Let's fix both.

asnan0830@gmail.com linkedin.com/in/abidiasnan github.com/stopabidi