A school built on the belief that AI education should be honest about what it teaches and what it does not

Cogniva was founded in 2019 by a small group of engineers and researchers who had spent the previous decade working on machine learning systems across the finance, logistics, and public sector domains in Malaysia and Singapore. The pattern they kept encountering was consistent: product teams and leadership understood AI poorly, and the available training materials either overstated what systems could do or skipped over the parts where they fail.

The founding decision was narrow in scope. Rather than building a broad technology school, Cogniva would offer three specific programmes — one for people who work near AI without building it, one for engineers who need to evaluate models rigorously, and one for organisations trying to introduce AI into existing processes sensibly. Nothing more ambitious than that.

The school opened its first cohort in George Town in early 2020. The short course ran to twelve participants in its first intake. The evaluation course followed in late 2020, and the consulting engagement was formalised in 2021 after several informal engagements with Penang-based technology firms. Since then the programmes have run in regular cycles, with cohort sizes kept deliberately small.

The location in George Town was chosen for its proximity to a working technology community in Penang's knowledge economy corridor and for the practical reason that the founding team was already there. All contact sessions take place in person. The consulting engagement can accommodate remote participation for specific components by arrangement.

Most working AI systems in 2025 are trained on large datasets, evaluated against held-out test sets, and deployed into processes where the consequences of errors range from minor inconvenience to significant operational or reputational harm. The people responsible for these systems — the engineers who build them, the product managers who specify them, the teams that use their outputs — often have very different understandings of where the risks sit.

Cogniva's programmes address this gap from three directions. The short course gives non-technical staff a grounded understanding of how models work and fail, drawn from published research and named governance frameworks. The evaluation course gives engineers practical methods for characterising model behaviour and documenting limitations, assessed against real work they bring to the course. The consulting engagement helps organisations build internal processes for evaluation and monitoring that do not depend on a single expert.

None of this is a substitute for the technical depth that comes from years of practice. We state that plainly. What the programmes offer is a structured, evidenced starting point — a well-kept notebook entry, not a finished thesis.

Cogniva is based in George Town because the founding team works here, and because Penang's technology and manufacturing sectors present a practical cross-section of the AI adoption challenges the programmes address. The school does not offer remote-only programmes; contact sessions are in person at the Jalan Sungai Pinang address.