Automation consultancy · Kuala Lumpur, Malaysia
The report someone rebuilds every morning. I make it build itself.
Someone on your team assembles the same report every morning, sorts the same inbox, retypes the same figures between two systems. I build the thing that does it instead — on your own hardware if the data can’t leave the building.
What I build, and what I build it with
Hover a rail to pause it · click a tile for detailWhat a real run looks like
Research Digest · daily 06:30 MYT06:30:02 fetch arxiv · 214 new papers 06:30:19 score · 214 → 11 above threshold 06:31:44 read pdfs · 11 full texts, local 06:36:10 summarise · ollama, on-device 06:38:02 validate · 11/11 pass rule set 06:38:25 typeset · digest.pdf, 6 pages 06:38:26 deliver · done. cloud spend RM0.00
Tell me the task you hate doing
Direct lineDescribe the manual job. I’ll come back inside 24 hours with a fixed quote and a plan — what gets automated, what stays human, and where it runs.
About
Five years inside regulated finance. Now I automate the work it left behind.
I’m Hakeem. I run a small automation consultancy out of Kuala Lumpur, working with Malaysian SMEs who are paying salaried people to do work a machine should be doing.
I did the five years first. BSc (Hons) Computer Science at Sunway University on the Jeffrey Cheah Continuing Scholarship, lab assistant at the House of Multimodal Evolution (H.O.M.E) Lab, then production systems in fintech and banking. At Curlec I built a subscription management system end to end in ReactJS, Django and PostgreSQL, dealing with the clients using it directly. At Aleph Labs I worked on UOB Bank internal systems as a ReactJS developer: secure authentication rolled out across seven countries, and the Profile Linkage Management System, both under strict security and compliance requirements. I’m now a full-stack engineer at Paywatch, where I led the FX system behind Genting Cruise’s seafarer payroll end to end — Node.js and Python APIs against third-party payment systems, PostgreSQL schema design, BigQuery reporting pipelines, and production n8n agent workflows running against live internal systems with human review checkpoints. That work teaches a specific set of habits: make every job idempotent so a re-run can’t double-charge anyone, treat authentication as something an auditor will read, handle errors loudly instead of swallowing them, and take only the data you actually need.
Those habits are the reason any of this is worth buying. An automation for a small business has to run at six in the morning with nobody watching, survive a supplier sending a malformed file, and stop rather than quietly write nonsense into your records. It also should not copy your customer data anywhere you didn’t agree to, which is why local models are the default here. Most of what I ship isn’t clever. It is unexciting code that keeps working, which is the only kind worth paying for.
“If the data can’t leave the building, the model doesn’t either.”
Selected work
Two systems that run without me.
Both replaced a recurring manual job. Both still running.