Teaching applied AI, one project at a time.

Tensorhaus started in 2021 as a small study group in a rented office near Sukhumvit. Three software engineers and a statistician were frustrated with the same problem: the online courses they could find were either too shallow to build real skills or too theoretical to be useful at work. They started meeting on Saturday mornings with a whiteboard, a stack of notebooks, and a shared drive of datasets.

The format worked. People came with specific questions about their own data and left with working code. By the end of 2022 the group had grown to include participants from several Bangkok companies, and the Saturday sessions were regularly full. The decision to formalise the curriculum into a structured school followed from that simple observation — the format produced better outcomes than watching lectures alone.

Tensorhaus opened its first intake in early 2023. The name was chosen to reflect the work done here: a tensor is the fundamental data structure in deep learning, and a haus is a workshop, a place where things are made. The school has kept that character. Lessons are practical. Feedback is written and specific. Projects are your own.

The team now includes six instructors and two programme coordinators, all of whom continue to work on applied AI projects outside of their teaching roles. That matters to us: the people giving feedback on your code are people who write similar code themselves.

Most people who want to learn machine learning are not students. They have jobs, schedules that change, and specific questions rooted in actual work they are doing or want to do. A programme that treats them as students in a classroom — with fixed attendance, graded exams, and long stretches of theory before any hands-on work — tends not to fit their situation well.

Tensorhaus builds around a different assumption. The core activity is the notebook. Each week's material is a Jupyter notebook with explanations woven into the code, not slides followed by an exercise. You read, run, observe, modify, and submit. The feedback you receive addresses your specific choices — why a particular approach does or does not hold up with your data.

The mentor sessions in the ML Pathway and Capstone add a layer that asynchronous feedback cannot replicate. Being able to share your screen, show someone a confusion matrix that does not make sense, and hear them think through it with you is different from reading a comment. We keep those sessions small enough that they remain actual conversations.

Bangkok has a growing community of people working in data-adjacent roles who want to go deeper. We see Tensorhaus as a place those people can do that work seriously, at a pace that fits their lives, with people who take their questions as seriously as they do.