A Practice Built
on Careful Work
Sovanta is an AI engineering practice based in Kuala Lumpur. We build systems that handle classification, voice, and data quality — with full attention to how they will actually be used.
← Back to HomeWhere Sovanta Comes From
Sovanta was founded in Kuala Lumpur with a straightforward premise: AI engineering work in Malaysia was either too generic or too expensive. Most organisations were offered off-the-shelf platforms that did not match their actual data environments, or they were quoted international agency prices for work that could be done locally.
The team behind Sovanta had spent years on machine learning projects for media companies, e-commerce platforms, and enterprise knowledge management systems across Southeast Asia. The decision to formalise this experience into a dedicated practice came from noticing the same problems recurring: poor data quality, unclear project scope, and deliverables that engineering teams could not maintain after handover.
Sovanta's service catalogue was shaped around three areas where well-scoped AI work produces clear and measurable outcomes — content classification, voice interface development, and data quality assessment. Each service has a defined duration, a transparent price, and a handover process.
Mission
To deliver AI engineering that Malaysian organisations can actually use — systems that are well-documented, maintainable, and aligned with the data realities of the region.
Values
- Honesty about what AI can and cannot do
- Scope discipline — delivering what was agreed
- Respectful client communication throughout
- Building for maintainability, not dependency
Focus Areas
Content classification at scale · Voice interface architecture · Data quality auditing and remediation planning · NLP pipeline development · Training data preparation
Our Team
Ahmad Razif
Lead ML Engineer
Eight years developing NLP and classification systems across media and e-commerce verticals. Leads model design and architecture at Sovanta.
Nurul Liyana
Voice & NLU Specialist
Focuses on intent parsing, wake-word detection, and voice user journey design. Previously worked on voice product features for a regional fintech platform.
Wei Kang
Data Quality Lead
Specialises in dataset auditing and remediation planning. Has assessed training data for ML readiness across manufacturing, logistics, and retail clients.
Standards & Protocols
Data Handling
All client data is handled under confidentiality agreement, stored on encrypted environments, and deleted or returned following project completion.
Defined Deliverables
Every project begins with a written brief that specifies deliverables, acceptance criteria, and timeline. Scope changes require written agreement from both parties.
Documented Code
All systems are delivered with technical documentation, inline comments, and a handover session. Clients receive everything needed to maintain the work independently.
PDPA Compliance
Sovanta's data practices align with Malaysia's Personal Data Protection Act (PDPA) 2010. We advise on relevant compliance considerations as part of each assessment project.
Milestone Reviews
Progress reviews at defined milestones ensure alignment before the next phase begins. Clients have visibility at each stage — no surprises at handover.
Evaluation-First Approach
Model performance is evaluated against agreed metrics before delivery. We document evaluation methodology so clients can reproduce assessments independently.
AI Engineering Expertise in Kuala Lumpur
Sovanta operates at the intersection of machine learning engineering and practical business application. The team's combined experience spans content intelligence, conversational AI, and data operations — disciplines that increasingly overlap as organisations begin building systems that understand, categorise, and respond to information.
Working from Kuala Lumpur places Sovanta at the centre of a growing regional AI services market. Malaysian organisations across sectors such as media, retail, logistics, and financial services are at varying stages of integrating machine learning into their operations — and the most common obstacle remains the same: data that has not been prepared or assessed for the purpose.
By combining data quality evaluation with downstream AI development services, Sovanta offers a coherent path from raw data to working system. Each engagement is short enough to produce tangible output, and well-documented enough to support internal continuity after handover.
Questions about how we work?
We are happy to explain our process in more detail before any commitment is made. A short call or email exchange is often enough to clarify whether a project is a good fit.
Reach Out to Us