Skip to content
The Builder Course
Back to all courses and cohorts

Agent design and product building

Build your AI agent online this September.

Cohort · September 2026

Turn one real workflow into an AI agent across five live weekday sessions in English and a live Defense Day.

View cohort details

What you leave with

A working, inspected AI agent.

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

This cohort

Everything you need to know before you join.

Online

Cohort · September 2026

24 seats remaining
Location
Live online, delivered in English, available worldwide.
8:00 PM – 10:00 PM GMT+8Your local time · detecting…. Sessions run 8:00 PM – 10:00 PM GMT plus 8.
Dates
Build: 15, 17, 18, 21 & 23 Sep
Defense: 30 Sep
8:00–10:00pm SGT
Commitment
Five 2-hour build sessions, plus Defense Day
Price
SGD 399
Per seat.
Introductory price
Powered by

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.

Seats are confirmed in payment order.

Weekend cohort interest

No tuition payment is due when you register interest.

Launch threshold · 8 paid participants

Weekend option

Weekend online

Live online

Four live online build sessions, three hours each, followed by Defense Day.

Schedule
19, 20, 26 & 27 September 2026 · 10:00am–1:00pm SGT · Defense Day 30 September
Commitment
Four 3-hour build sessions, plus Defense Day
Price
SGD 399 per seat
Where
Live online in English · available worldwide
Powered by

If this cohort runs, includes US$100 Codex + US$50 API credits.

Register your interest in this schedule in the next step.

What is an agent?

An agent you deploy, connect, and conduct.

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

    • Skills, the right tools (MCPs), and a memory system.
    • Permissions and boundaries, deployed to run in the cloud.
    • Keeps working even when you're away from your desk.
    • Evals watch its quality: true to your instructions, or does it need more tools or inputs?
    • You conduct it toward a goal: set direction, read the evals, adjust. You don't operate it.

    You conduct

    Set the goal · read the evals

    Agent

    Running in the cloud

    SkillsTools · MCPsMemoryBoundaries

    Evals

    On track? Needs more tools or inputs?

    Read the evals, adjust, and it keeps running.

    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

See how the work develops.

Each phase advances the same practical piece of work, so the course reads as a progression rather than a list of topics.

  1. 01

    Pre-work

    Choose the workflow and get literate.

    Select a workflow worth turning into an agent and establish the product, agent, and design foundations needed to build it.

    • Agent literacy
    • Workflow selection
    • Design literacy
    • Tooling setup
  2. 02

    Build Day 1

    Give the agent clear capabilities and limits.

    Write the workflow boundary, then assemble skills, tools, memory, approval points, and the interface foundations around it.

    • Workflow scope
    • Skills and tools
    • Memory and boundaries
    • Design foundations
  3. 03

    Build Day 2

    Make the build observable.

    Deploy the reference pattern and use evals to inspect quality, cost, security, and drift instead of trusting a polished demo.

    • Cloud deployment
    • Evals
    • Cost and security
    • Observability and drift
  4. 04

    Async build period

    Build the agent for your own workflow.

    Reuse the reference pattern on your own problem, with peer feedback and support while you work through the edges.

    • Your workflow
    • Full agent build
    • Peer feedback
    • Iteration
  5. 05

    Defense Day

    Explain the consequential choices.

    Present the agent and its design system, then defend the scope, behavior, architecture, safety, cost, and evidence behind it.

    • Product scope
    • Design coherence
    • Agent behavior
    • Reliability and trade-offs

Private group context, schedule, and delivery needs can be adapted while keeping the five-stage course arc.

What you will walk away with

What you keep.

You leave with one working agent, one design system, and a reusable way to build the next one.

  1. A reusable build process

    A simple way to choose a workflow, set boundaries, test behavior, and improve the result.

    • Choose the right workflow
    • Set rules and success criteria
    • Review what happened and improve it

  2. A working AI agent

    One agent for your own work, built from the reference pattern and safe to keep using.

    • Clear inputs and outputs
    • Allowed tools or data
    • Human approval where it matters

  3. A design system

    A small visual system so AI-built screens feel like one product.

    • Tokens, type, color, spacing, and components
    • States for waiting, failing, and recovering
    • Reusable patterns for future AI-built work

Defense Day

Defend the decisions, not the demo.

  1. Workflow scope

    Did you choose a workflow worth turning into an agent?

  2. Design coherence

    Does the interface feel like one usable product?

  3. Agent behavior

    Are the instructions, tools, constraints, and approval points clear?

  4. Architecture and reliability

    Will the workflow survive beyond the first polished demo?

  5. Cost and security

    Can the workflow run without surprise spend or unsafe access?

  6. Observability and drift

    Can you see what happened, what changed, and how you would correct it?

Your facilitators

Practitioners who teach, then press the work.

Design, product, and engineering practitioners teach the core judgment layers and return for Defense Day.

Deepika Murthy

Deepika Murthy

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

  • Scope
  • Judgment
  • Trade-offs
  • Commercial logic
Rajat Goyal

Rajat Goyal

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.

Previously

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

  • Production-readiness
  • Architecture
  • Cost & security
  • Observability & drift
Keith Oh

Keith Oh

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

  • Taste
  • Craft
  • User clarity
  • Interaction logic
  • Design trade-offs
Pritha Vijay

Pritha Vijay

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.

Previously

Teaches

Product fundamentals: scoping what's worth building, prioritizing ruthlessly, and making the call when the evidence is incomplete.

Presses on Defense Day

  • Scope
  • Prioritization
  • Problem definition
  • Trade-offs

The room helps round out whichever layer you are stretched on.

Who this is for

For people already building with AI.

For PMs, product leaders, founders, operators, designers, and engineers already using AI in real product work.

You will fit if you…

  • Already use AI to draft, code, summarize, analyze, or prototype
  • Want to turn one real workflow into an agent
  • Want a guided build-along before building your own
  • Can generate outputs, but want sharper judgment
  • Want AI-built work to feel coherent

What this is not

  • Not a coding bootcamp. AI does the typing, but you must understand what it produced.
  • Not prompt-engineering training. Prompts are part of the work, not the whole course.
  • Not a Figma course. Design fundamentals are taught for judgment, not tool training.
  • Not for production enterprise integrations, real employer data, or uncontrolled write-backs.

Support & community

You're not on your own, during the cohort or after.

The agent and design system are what you build. The room is what you keep.

Async support, all week

The team stays in the group channels through the week, not just on live days.

Optional live AMAs

When chat is not enough, optional live AMAs unblock the room.

The builders you keep

The room is what you keep.

FAQ

Questions, answered plainly.

The answers here apply to this exact cohort.

I already ship products. Will this be too basic?

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.

My engineering is weak. Will I be lost?

No. AI does the typing. We teach the engineering concepts in plain terms so you can make the calls and explain them.

Do I need an agent idea before Day 1?

Helpful, not required. Pre-work helps you find candidate workflows, and Day 1 pressure-tests the scope.

What do I actually own, and can I use it commercially?

The design system and agent you build are yours to keep and use in your own work.

Can I join the online AI agent course from outside Singapore?

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.

Is this an LLM course or an AI agent building course?

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.

Is this an AI certification course?

No. This is a practical cohort course focused on producing and defending a working agent, not an accredited certification or professional qualification.

Does the course cover AI transformation?

The course teaches a repeatable way to transform one real workflow with AI. It is not an enterprise-wide transformation or consulting program.

What happens to my S$10 priority deposit?

If your deposit is paid, use the same registration email and it will be credited automatically at checkout, even if you originally selected another September or October format. Without another discount, your remaining tuition is SGD 389.

Spread the word

Know someone who builds with AI and sweats the details?

Cohort · September 2026 · Build: 15, 17, 18, 21 & 23 Sep · Defense: 30 Sep · Live online in English · available worldwide. Send it to the practitioner who should be in the room.

LinkedInPost on X

Cohort · September 2026

Build: 15, 17, 18, 21 & 23 Sep · Five 2-hour build sessions, plus Defense Day · SGD 399