Agent design and product building
Build with AI. Judge with taste.
A live, practical AI agent building course for people who want to turn one real workflow into an agent they can inspect, improve, and explain.
Choose an English-language online cohort from anywhere in the world, or join an in-person cohort in Singapore. Online session times are shown in your local timezone.
What you leave with
Build one agent for your own work, the design system around it, and a reusable process for scoping, testing, and improving the next one.
Bring one real workflow
Named cohorts
Every cohort is a small room on purpose: enough attention on your build to keep you moving, and enough builders beside you to keep it honest. Choose the dates and format that fit.
Includes US$100 Codex + US$50 API credits.
We're with you on the community group throughout, plus remote unblock sessions whenever you're stuck.
We're with you on the community group throughout, plus remote unblock sessions whenever you're stuck.
Seats are confirmed in payment order.
We're with you on the community group throughout, plus remote unblock sessions whenever you're stuck.
Seats are confirmed in payment order.
There's a ladder from prompting to a real agent. Each rung hands off more of the work, until it runs without you.
You operate a prompt, a skill, a connected workflow. You conduct an agent: deployed to the cloud with its own tools, memory, boundaries, and evals. It keeps working, and stays on track, even when you're away from your desk.
From prompt to agent
You conduct
Set the goal · read the evals
Agent
Running in the cloud
Evals
On track? Needs more tools or inputs?
Deployed to the cloud, it keeps working when you're away, and evals tell you when it needs more tools, inputs, or direction.
Curriculum and delivery
Each phase advances the same practical piece of work, so the course reads as a progression rather than a list of topics.
Pre-work
Select a workflow worth turning into an agent and establish the product, agent, and design foundations needed to build it.
Build Day 1
Write the workflow boundary, then assemble skills, tools, memory, approval points, and the interface foundations around it.
Build Day 2
Deploy the reference pattern and use evals to inspect quality, cost, security, and drift instead of trusting a polished demo.
Async build period
Reuse the reference pattern on your own problem, with peer feedback and support while you work through the edges.
Defense Day
Present the agent and its design system, then defend the scope, behavior, architecture, safety, cost, and evidence behind it.
Private group context, schedule, and delivery needs can be adapted while keeping the five-stage course arc.
What you will walk away with
You leave with one working agent, one design system, and a reusable way to build the next one.
A simple way to choose a workflow, set boundaries, test behavior, and improve the result.
One agent for your own work, built from the reference pattern and safe to keep using.
A small visual system so AI-built screens feel like one product.
Defense Day
Did you choose a workflow worth turning into an agent?
Does the interface feel like one usable product?
Are the instructions, tools, constraints, and approval points clear?
Will the workflow survive beyond the first polished demo?
Can the workflow run without surprise spend or unsafe access?
Can you see what happened, what changed, and how you would correct it?
Design, product, and engineering practitioners teach the core judgment layers and return for Defense Day.
Lead instructor · Product judgment & AI product building
Nineteen years across engineering, growth, marketplaces, platforms, and product. At Gojek, led product strategy for pricing, matching, batching, allocation, and fulfillment systems across Transport, Food, and Logistics, while driving GenAI adoption across the PM organisation. Earlier at Rakuten Viki, moved from video engineering into product and growth, launching and scaling a global subscription product to 200,000+ paid subscribers. Co-founder of Women in Product Singapore.
Previously
Teaches
Product scoping, prioritization, AI product judgment, and structured defense: deciding what is worth building, what evidence is enough, and how to defend the trade-offs.
Presses on Defense Day
Applied AI and Agentic products
A builder across fintech, edtech, consumer apps, ride-hailing, and applied AI, with a decade of building and scaling production systems and the teams behind them. Most recently leading learning systems and applied AI at Aampe, scaling millions of per-user AI agents that personalize 1-to-1 and building agentic martech products. Earlier headed transport engineering at Gojek (millions of daily bookings across Singapore and Indonesia), and built financial and education platforms at Arcesium (D. E. Shaw Group) and Vedantu. Joins the Defense Day panel.
Teaches
The engineering in plain terms (architecture, observability, cost, security, and drift), and the judgment to tell when an agent's ready to ship.
Presses on Defense Day
Product design
Product and design leader with experience across consulting and in-house roles, shaping digital products across markets, cultures, and organisational contexts. Most recently led Product Design and Core Marketplace product at Carousell. Earlier roles span IDEO, Rakuten Viki, and DSTA, combining product strategy, design craft, research, systems thinking, and human-centred problem solving.
As Honorary Secretary of Design Business Chamber Singapore, he champions better business by design, and contributes to conversations on how AI is changing design, product, and creative practice. His background spans Electrical & Computer Engineering, Psychology, and Human-Computer Interaction.
Previously
Teaches
Design basics, critique, judgment, usability, and taste, plus how these help turn rough AI outputs into thoughtful product experiences.
Presses on Defense Day
Product Management
Sixteen years building and scaling products across AI, marketplaces, and consumer platforms. At Uber, built the AI Solutions Marketplace, a human-in-the-loop platform for enterprise AI model development, from the ground up, owning supply, matching, dispatch, payments, and integrity end to end. Earlier at Grab, led product across safety, onboarding, and fraud for 180 million users across eight markets through IPO. Began her career in engineering and general management across six countries before moving into product through Strategy& and a stint in robotics. Studied engineering at IIT Kanpur and earned her MBA from Harvard Business School.
Teaches
Product fundamentals: scoping what's worth building, prioritizing ruthlessly, and making the call when the evidence is incomplete.
Presses on Defense Day
The room helps round out whichever layer you are stretched on.
For PMs, product leaders, founders, operators, designers, and engineers already using AI in real product work.
Support & community
The agent and design system are what you build. The room is what you keep.
The team stays in the group channels through the week, not just on live days.
When chat is not enough, optional live AMAs unblock the room.
The room is what you keep.
FAQ
Anything else, ask us before you enroll.
If you can already build, test, and explain agents you keep using, you're past this. If you ship with AI but want more reliable judgment, this is for you.
No. AI does the typing. We teach the engineering concepts in plain terms so you can make the calls and explain them.
Helpful, not required. Pre-work helps you find candidate workflows, and Day 1 pressure-tests the scope.
The design system and agent you build are yours to keep and use in your own work.
Yes. Online cohorts are taught live in English and are open worldwide. Builders can join from Australia and New Zealand; Singapore, Thailand, Cambodia, and the rest of Asia-Pacific; Germany and wider Europe; Dubai, the UAE, and the Middle East; the United States, Canada, Africa, and South America. Check the cohort dates and the local session time shown on this page before enrolling.
It is a hands-on AI agent building course. You use large language models as part of a deployed agent system and learn tools, memory, boundaries, evaluations, and human approval points. It is not a machine-learning degree or a course in training foundation models.
No. This is a practical cohort course focused on producing and defending a working agent, not an accredited certification or professional qualification.
The course teaches a repeatable way to transform one real workflow with AI. It is not an enterprise-wide transformation or consulting program.
Turn one real workflow into an AI agent
Cohort 1 · Singapore · Build: 25 Jul & 1 Aug · SGD 499
