Core 01 — AI solutions
AI that does real work.
Assistants that answer from your own documents, agents that take orders on WhatsApp, and vision models that read what cameras see. Built for production — or fully offline on your own hardware.
Odoonic — software studio, Islamabad
Software,engineered.
AI solutions, Odoo ERP and websites — designed, built and looked after by one senior team.
One studio. Four cores.
Every Odoonic project runs on the same four disciplines. Use one — or let them work together.
Core 02 — Odoo ERP
Odoo, set up right.
From first install to custom modules and version upgrades. We configure the apps you need, connect the tools you already use, and leave your team trained and confident.
Core 03 — Web development
Websites people remember.
Fast, accessible sites and web apps with craft in the details — from a sharp landing page to a SaaS MVP, with interactive 3D where it earns its place.
Core 04 — Website care
Kept fast, secure and current.
Monthly care plans so nothing drifts: we watch uptime, patch and update, back everything up, and make the small edits you'd otherwise put off.
Better together.
An Odoo rollout with AI built in. A website that feeds your CRM. A care plan that keeps all of it healthy.
Odoonic is a small studio of engineers. We build AI that does real work, set Odoo up the right way and design websites people remember — then we stay to keep it all running.
Private AI
192GB
of VRAM in our own lab — the machine we use to size, test and tune private AI before it runs on yours.
Your data never leaves your servers.
For finance, health and government work we deploy open-weight models on hardware you control — sized, quantised and tested on the same kind of machine we run ourselves.
- GPUs
- 2192 GB VRAM combined
- Models
- Open-weightDeepSeek-R1 · Kimi K2 · Llama
- Runtime
- llama.cppOllama · GGUF quantisation
- Network
- OfflineNothing leaves the building
Selected work
Built, shipped, still running.
Client work, open source and our own infrastructure — each one in use today.
Assalam o alaikum — 2 chicken karahi aur 6 naan bhej dein
▶0:07
Order confirmed
2 × Chicken Karahi · 6 × Naan
Sent to the kitchen. Shukriya!
01 / 04In production
AI agents
WhatsApp Ordering Bot
Customers order by chat or voice note — in Roman Urdu — and confirmed orders post straight to the shop's order system. One deployment serves many shops.
- WhatsApp Cloud API
- LLM
- Speech-to-text
- Python
- REST
# offline · 2 GPUs · 192 GB VRAM $ llama-server -m deepseek-r1.gguf \ --n-cpu-moe 20 -c 32768 model loaded ✓ listening on 127.0.0.1:8080 egress: none — data stays on-prem
04 / 04Internal
AI infrastructure
Private LLM infrastructure
DeepSeek-R1 and Kimi K2–class models running fully offline on a dual-GPU, 192 GB VRAM machine — the setup we use to size private AI for clients.
- llama.cpp
- Ollama
- GGUF
- MoE offloading
How we work
A working demo, every week.
No long silences and no surprises at launch — you see the thing being built.
- 01
Discover
One call and a written brief: goals, current systems, constraints, budget.
- 02
Design
Architecture, flows and visual direction you can react to before we build.
- 03
Build
Short cycles with a working demo every week. Progress you can click, not status reports.
- 04
Launch
Data migration, training and a calm go-live with monitoring switched on.
- 05
Care
Updates, backups, fixes and improvements on a monthly plan.
The studio
Small on purpose. Senior by default.
Run from Islamabad by the people who write the code. We take on a few projects at a time, so each one gets both of us.
Muhammad Soban Hameed
Co-founder · AI engineering
LLM apps and agents, computer vision with YOLO, Odoo AI modules, FastAPI services.
Syed Hasnain
Co-founder · AI/ML & full-stack
Machine learning, React and Next.js front ends, cloud on AWS and Azure, mobile.
01
One team, build to care
The people who build your system are the people who look after it. No hand-offs, no lost context.
02
AI we actually run
Large models run on our own 192 GB VRAM machine, so we know what private AI costs, how fast it is and where it breaks.
03
Straight to the engineers
You talk to the founders doing the work — on a call or on WhatsApp — not to an account manager.
Start a project
Let's build what's next.
A short call is usually enough to tell whether we're the right fit — and what it would take.
Book a call
hello@odoonic.com