Our Company
Engineering discipline applied to machine learning
Wira Intelligence was founded on the premise that most AI projects fail not because the underlying models are wrong, but because the surrounding engineering is underdeveloped.
← Back to HomeOur Story
From prototype to something your team can rely on
Wira Intelligence was established in Kuala Lumpur after years of observing a recurring pattern in the Malaysian technology sector: internal teams produce ML prototypes that perform well in controlled conditions, then encounter production realities — data drift, infrastructure constraints, lack of monitoring — that the prototype was never designed to handle.
The founding team came from engineering backgrounds rather than research backgrounds. That distinction shapes how we approach every engagement. We are not primarily interested in novel model architectures or state-of-the-art benchmark scores. We are interested in systems that work reliably in the specific environment your organisation operates in, and that your team can maintain after we leave.
We keep our client list deliberately small. Taking on more work than we can do thoroughly is the one thing that would undermine the only thing we are selling: careful, attentive engineering practice.
Our Mission
What we are trying to do
"Help Malaysian organisations build AI systems that hold up — not just in a notebook, but in the hands of the people who depend on them."
We take on three types of work: engineering ML prototypes to production standard, reworking retrieval and search pipelines that have underperformed, and providing ongoing mentorship to internal teams building AI capability for the first time.
In each case, we measure what we do. Engagements without explicit success criteria and reproducible evaluation methods are not something we take on. A system that cannot be measured cannot be said to have improved.
The Team
The people doing the work
Ahmad Haikal
Principal Engineer
Leads production engineering engagements. Came to AI work from a background in distributed systems, which informs his view that reliability engineering and ML engineering are the same discipline applied to different components.
Nurul Rashidah
Search & Retrieval Specialist
Leads retrieval re-engineering engagements. Has spent several years working specifically on the problems that arise when retrieval systems encounter multilingual corpora — a common situation in Malaysian organisations.
Siew Lim
Mentorship Lead
Leads team mentorship engagements. Works directly with engineers each week, reviewing actual code and system decisions rather than providing abstract guidance. Was previously a senior data engineer at a Kuala Lumpur fintech firm.
How We Work
Standards we apply to every engagement
Written Scope Before Work Begins
No engagement starts without a written scope document agreed by both parties. This defines what will be built, how success will be measured, and what is explicitly out of scope.
Evaluation Methodology as Output
Every engagement delivers not just the improved system but the method used to measure improvement. Your team can rerun the evaluation at any point without our involvement.
Confidentiality as Standard
A confidentiality agreement is in place before any client data or system details are shared with us. We do not discuss client systems with third parties and do not retain client data after an engagement concludes.
Documentation as a Deliverable
Written documentation covering design decisions, known limitations, and operational guidance is part of every deliverable. A system without documentation places an unfair dependency on the people who built it.
Collaborative Engagement Structure
We work alongside your engineers, not separately from them. Each engagement includes structured knowledge transfer so your team understands what was built and why the design decisions were made.
Honest Scope Limitation
We decline work that falls outside our defined engagement types, and we will tell you if we believe your situation requires a different kind of assistance. We do not take on work in order to figure out how to do it.
Expertise
The technical areas we work in
Wira Intelligence operates in three technical areas: ML systems production engineering, retrieval and search system re-engineering, and AI engineering mentorship for internal teams. These three areas were chosen because they represent the most common gaps we encountered in Malaysian organisations attempting to operationalise AI work — not because they are the most fashionable or fastest-moving areas of the field.
Production engineering for ML systems requires a different set of concerns than prototype development. The key questions shift from "does this model produce correct outputs on test data?" to "how does this system behave when input data shifts over time?", "how does the team know when outputs have degraded?", and "what is the path back to a working system when something goes wrong?" These are engineering questions, not research questions, and they require engineering answers.
Retrieval system quality is often poorly measured. Organisations frequently evaluate search quality using proxy metrics — click rates, session depth — rather than direct evaluation of retrieval accuracy against the actual information needs of the people using the system. Our retrieval engagements begin by establishing what good retrieval would look like in your specific context, and then measuring the current system against that definition before any changes are made.
Engineering mentorship for AI work differs from conventional training in that it is structured around the actual work the team is doing at the time, not a curriculum designed in advance. The weekly review sessions examine real code, real design decisions, and real production incidents, with a focus on developing judgment rather than knowledge of specific tools or frameworks.