Depth of knowledge applied with care
Tensora is an AI consulting practice built around the belief that useful AI comes from understanding data honestly — not from applying templates quickly.
Back to HomeStarted with a disagreement about what AI could do
Tensora came out of a recurring frustration: organisations were being told that AI would transform their operations, but the actual work of preparing data, defining what success looked like, and maintaining models after deployment was consistently underestimated or ignored.
We started as a small group of practitioners in Singapore who had each spent years working on the applied side of data science — building forecasting systems for financial institutions, designing conversational tools for service organisations, and connecting fragmented knowledge bases into coherent information architectures.
The practice is named after tensors — the multi-dimensional structures that underlie most modern AI computation. It reflects a preference for precision in how we talk about the technology and what it can achieve in a given context.
We work with organisations across Singapore that are moving from curiosity about AI toward specific, deployable capabilities. Our engagements are scoped to be realistic, our documentation is thorough, and we stay involved until the handover is clean.
Our Mission
To help organisations in Singapore deploy AI that works reliably within their actual constraints — built on their data, integrated with their systems, and understood by their teams.
Our Approach
Every engagement begins with a frank data assessment. We do not proceed with build phases until we are confident the data foundation supports the intended outcome. This avoids wasted investment and misaligned expectations.
Our Commitment
We transfer full ownership of all deliverables to clients. Documentation, code, and model artifacts belong to you — along with enough explanation to maintain and extend what we built.
Practitioners, not account managers
Kai Hartono
Principal, AI Systems
Twelve years in applied machine learning across financial services and logistics. Leads forecasting and knowledge graph engagements with a focus on interpretability and production stability.
Serena Lim
Lead, Conversational AI
Specialises in dialogue design and NLP system architecture. Has built conversational systems for healthcare, government services, and retail across Southeast Asia.
Rohan Nair
Knowledge Engineering
Background in ontology design and semantic web technologies. Leads knowledge graph construction engagements with expertise in entity extraction and relationship modelling at scale.
How we maintain quality across every engagement
These are the specific practices that shape how we assess, build, and hand over AI work.
Data Privacy Compliance
All engagements adhere to Singapore's Personal Data Protection Act. Data handling procedures are documented and agreed before any information is shared.
Version-Controlled Deliverables
All code, models, and configuration files are managed in version control repositories. Clients receive access to the full history of changes and the rationale behind key decisions.
Structured Testing Protocols
Models and systems are evaluated against defined performance criteria before handover. Testing coverage and results are included in all delivery documentation.
Assumption Documentation
Every model includes a written record of the assumptions made during design, the data conditions it was built for, and the circumstances under which its outputs should not be trusted.
Mutual Confidentiality
A mutual non-disclosure agreement is signed before any engagement begins. Client data and business context are never shared or repurposed outside the specific project scope.
Knowledge Transfer Sessions
Each engagement concludes with dedicated sessions to walk your technical team through the system, explain design decisions, and address questions before sign-off.
What we bring to AI engagements in Singapore
The organisations we work with most often are those that have accumulated substantial data — transaction records, operational logs, communications, structured databases — and are starting to ask how AI might be applied to that material in a way that produces reliable, interpretable outputs rather than opaque black boxes.
Time series forecasting, for instance, is most useful when the practitioner understands both the mathematical requirements of the chosen model and the domain context of the data — why certain periods show anomalies, what external factors might explain seasonal patterns, and where the limits of prediction genuinely lie. This kind of combined understanding is what we bring to scoping and execution.
In conversational AI, the difference between a system that frustrates users and one that serves them well often comes down to careful dialogue architecture — how ambiguous inputs are handled, when the assistant should escalate to a human, and how domain boundaries are respected. These decisions happen in design, not after deployment.
Knowledge graph work requires comfort with both technical ontology design and the messier reality of how knowledge is actually distributed across an organisation's documents and databases. The value is in the connections between entities as much as the entities themselves, and the quality of the resulting graph depends on how carefully those connections are mapped and validated.
Working within Singapore's business environment also means familiarity with the relevant regulatory context — the PDPA, MAS guidelines where applicable, and the norms of sectors that are significant in the Singapore economy including financial services, logistics, and professional services.
Find out if there is a sensible fit
A short initial conversation about your data situation and what you are trying to achieve is enough for us to give an honest view of what is feasible.
Start a Conversation