Services
Automation & integrations for small businesses
If your team copies data between systems, chases the same updates every week, or wants AI without the guesswork, I build the piece that does it for them. Small, well-tested tools that fit around the software you already use.
What I can help with
- integrations
Integrations & data migration
Get the tools you already pay for talking to each other, or move years of records from one system to another without losing the history.
- Sync a helpdesk, project tracker, CRM or spreadsheet so nobody re-types data
- Move tickets, comments and attachments between platforms
- Connect payments, billing or sign-ups to the rest of your stack
- apis · webhooks
APIs, webhooks & automations
Small, reliable services that react when something happens: a form is sent, a payment lands, a device goes quiet. They run on their own and tell the right person.
- Webhook receivers that don't drop events, with a log you can replay
- Scheduled checks that send an alert to Teams or Telegram when something needs a person
- Internal tools inside Microsoft Teams, signed in with your Microsoft accounts
- llm · rag
AI assistants on your own documents
A question-answering assistant that reads your policies, manuals or past tickets and answers from them, and says plainly when the answer isn't there.
- Internal Q&A over handbooks, procedures or product docs
- Drafting replies or summaries from your own records
- Testing that answers stay grounded before anyone relies on them
- Built on Azure AI (Microsoft certified, AI-103) or the Claude API
Built before:RAG Autopsy - ai · guardrails
Adding AI to a product, safely
Put an LLM into an existing app or workflow with the guardrails in place: protection against prompt injection, and customer data handled under a clear privacy model.
- Classify, extract or clean up incoming data with an LLM
- Screen user input for prompt-injection attacks
- Keep sensitive data out of logs and model providers
Case study · in my role at Energy One
Moving 10,000+ support tickets to a new platform
The problem
Customer support history across dozens of organisations had to move from Zendesk to Jira Service Management. Each ticket carried its comments, attachments and customer accounts, and some had more than a hundred of each. Copying that by hand was never an option.
What I did
I scoped, built and evolved major parts of the Python tooling that moved it, working with support, product, QA and operations colleagues on how each field should map. The tooling kept track of its own progress, so an interrupted run picked up where it left off, and it moved up to six tickets at a time instead of one.
- support tickets in scope
- 10,000+support tickets in scope
- migration success
- ~99.8%migration success
- throughput of the first version
- ~5.4×throughput of the first version
At the original one-at-a-time rate, the job would have taken days, not hours. Failures were logged for follow-up rather than stopping the run.
Read the full engineering write-up →How it works
- 01
A 15-minute call
You describe the manual work or the broken handoff. I ask questions.
- 02
A written scope
What I'd build, what it won't do, and a fixed quote before anything starts.
- 03
Build and hand over
Tested, deployed, and documented so it isn't a black box.
Got a process that eats someone’s week?
Tell me about it in 15 minutes. If it isn’t worth automating, I’ll say so.
Book a 15-minute callor email jobs@usmanmateen.com