We build AI that
runs in production
Thoughtium is an applied AI consultancy based in Kuala Lumpur. We work with organisations across supply chain, platforms, and software engineering to move machine learning models from concept into operational use.
Back to HomeBuilt from a frustration with AI that never ships
Thoughtium came out of conversations between practitioners who had seen the same pattern repeat across different companies: substantial investment in ML research, prototypes that performed well in evaluation, and then — nothing. Models that sat in notebooks while the business carried on without them.
The team behind Thoughtium had worked across logistics firms, digital platforms, and software companies throughout Malaysia and the broader region. The gaps were consistent: no clear scope before a project began, no infrastructure to operationalise models, and no handover process to keep things running after an engagement ended.
We set up in Kuala Lumpur to address that directly. Every service we offer is built around a tangible deliverable — a running model, an operational pipeline, a system you can hand to your team and maintain. The goal has always been AI that actually integrates into how your organisation works, not AI as a research exercise.
Years of ML practice
across the team collectively
Completed engagements
across Malaysia and Southeast Asia
Core services
each with a defined scope and deliverable
Mission and working principles
Honesty about feasibility
We won't take on work that isn't feasible or where we don't see a path to a useful outcome. If your data isn't ready, or the problem scope is unclear, we'll tell you before any contract is signed.
Written scope, always
Every project begins with a concise scope document both sides have agreed on. This removes ambiguity about what we're building and how we'll know when it's done.
Handover that actually works
A model your team can't operate isn't a deliverable. Documentation, a handover session, and 30 days of post-delivery support come with every engagement.
Transparency at every stage
We communicate openly during projects — including when something isn't going as expected. Clients don't get surprises at the end of an engagement.
Fit over novelty
We choose approaches that fit the problem, not approaches that are fashionable. If a simpler model solves your problem well, that's what we'll recommend.
Data handled with care
NDA before any data is shared. Client data is used exclusively for the engagement and deleted after delivery unless otherwise arranged. We operate under Malaysia's PDPA 2010.
People behind the work
Reza Amirul
Co-founder & ML Engineer
Leads technical delivery across supply chain and MLOps engagements. Previously built production ML systems for e-commerce and logistics firms across the region.
Nurul Farhana
Co-founder & Data Scientist
Focuses on model development and evaluation, particularly for content moderation and classification problems. Background in NLP and multimodal systems.
Leonard Khoo
Infrastructure Engineer
Responsible for deployment architecture, CI/CD pipelines, and monitoring setup on MLOps engagements. Works across AWS, GCP, and Azure environments.
Quality standards across every engagement
Version-controlled deliverables
All code, model artefacts, and configuration are version-controlled and handed over in a repository you own. Nothing is delivered as an opaque binary.
Tested and evaluated
Models are evaluated against agreed metrics before delivery. We document evaluation methodology and results so you have a clear baseline for future monitoring.
Privacy and data compliance
All engagements are conducted under NDA. We comply with Malaysia's Personal Data Protection Act 2010 and can accommodate additional data handling requirements on request.
Documented thoroughly
System documentation covers architecture decisions, data schemas, model behaviour, and operational runbooks. Your team can onboard without needing us in the room.
Observable in production
Where applicable, we configure logging, metrics, and alerting as part of delivery. Models in production need to be monitored — we build that in from the start.
Regular communication
Check-ins are scheduled at meaningful intervals, not just at the start and end. You'll know what's happening throughout the engagement, not just when we're done.
AI practice in the Malaysian and Southeast Asian context
Malaysia's data economy has grown substantially over the past decade, and ML adoption across logistics, finance, and platform businesses has accelerated alongside it. Thoughtium works within this context — understanding local cloud infrastructure, data residency considerations under the PDPA, and the practical realities of development teams in the region.
Supply chain operations across Malaysia often involve complex multi-tier networks spanning peninsular and East Malaysian locations, as well as cross-border flows into Thailand, Singapore, and Indonesia. Our modelling work accounts for these structural realities rather than applying generic frameworks that assume uniform infrastructure.
Content moderation requirements in Southeast Asia also carry regional specificity — language diversity, platform-specific policy contexts, and the need to handle Bahasa Malaysia, English, and Mandarin content within the same moderation pipeline. Our taxonomy and training data work reflects this.
The MLOps engagements we deliver are designed for engineering teams that often work with a mix of cloud environments and on-premise infrastructure. We build pipelines that can operate within those constraints, rather than assuming a greenfield cloud-native setup that doesn't reflect how most organisations here actually work.
Start with a conversation about your situation
We're happy to discuss what you're working on before any commitments are made. A scoping call takes about an hour and leaves you with a clear sense of what's possible.
Get in Touch