AI tools & automation in Malaysia
Start with a useful task, then decide where AI belongs. I build tools that connect models, data and interfaces into a workflow people can actually use.
Freelance development by Hakimi Sa’ari, based in Selangor and working with clients across Malaysia.
Who this is for
For businesses and teams exploring a specific automation idea, or trying to turn a working AI experiment into a more usable application.
What we could build
- Audio transcription and review workflows
- Document extraction, classification and search
- Computer vision collection and review tools
- API integrations that reduce repetitive steps
What the work can include
- A defined use case and a practical prototype where appropriate
- Model or API integration with a usable interface
- Error handling, review steps and evaluation against agreed examples
- A clear discussion of running costs, data flow and limitations
Before we start
AI outputs can be wrong. Important results need appropriate review, and model performance must be evaluated using relevant examples. API fees and third-party data processing are discussed before selecting providers. Portfolio demo metrics are not client results.
Related personal work
These are personal projects, not client commissions. Some screenshots use clearly labeled demo data. The write-ups explain exactly what I built.
- Offline Transcriber — Record or upload Malay/English audio, then review timestamped transcripts and speaker labels.
- Computer Vision Workbench — Manage datasets, inspect data quality, train models, and review predictions in one workspace.
- Exposure — Plan, generate, compare, and refine AI image edits in one local workspace.
How it works
Tell me about the work
Email the problem, the intended users and any examples. You don’t need a complete technical specification.
Agree the scope
We clarify features, constraints, deliverables and what is outside scope. I’ll prepare a quotation and proposed milestones around that discussion.
Build and review
Development follows the agreed milestones, with opportunities to review progress and test the main workflows.
Test and hand over
We work through acceptance checks, deployment or setup, and the agreed documentation. Maintenance and future changes are scoped separately.
Questions you might have
Can everything run locally?
Some workloads can use local models, subject to hardware and performance needs. Others suit hosted APIs. We’ll compare the data requirements, costs and practical limits before choosing.
Can you guarantee the AI will always be correct?
No. We should agree how to evaluate the feature, what a user must review and what happens when an output is uncertain or unavailable.
Let’s work together
Tell me what you need the software to do. We can work out the scope, the approach and what a useful first version should include.