First session Sat 3 October 2026
Applied AI Bootcamp — starts 3 October
Apply AI at full power in your day-to-day work. No coding needed.
12/15 Seats left
Cohort DetailsApplied AI Bootcamp — starts 3 October
Apply AI at full power in your day-to-day work. No coding needed.
12/15 Seats left
Cohort DetailsFor developers who already code. Sundays: Google Cloud mornings or Microsoft evenings.
Google Cloud in the mornings, Microsoft in the evenings. Each track is its own cohort of twenty, with the same screened entry, in person at Hatch Works, Colombo.
Real BuildrLabs sessions at Hatch Works, Colombo, September 2026: working developers at their laptops with a practitioner at the table. This is the room your cohort builds in.
Thirty two contact hours in person, under LKR 3,000 per hour, with Google Cloud credits and every lab included in the fee.
Guided lab, then unguided lab, then your own build. That progression is what turns a workshop into a bootcamp.
Last week’s pull requests on screen, learners present
The week’s core idea, taught against the system being built
Instructor at the front, everyone at the same checkpoint by the end
Spec given, learners build, instructor and TA float
This week’s ship defined, homework briefed
Pick an enterprise vertical on day one and keep it throughout. Synthetic data and mock APIs are supplied, so you build against something that behaves like a real enterprise.
Policy control, injection defence
Workflow graph, deterministic paths
Grounded on enterprise data
Tools and MCP, human approval gate
Remote, reached over A2A
Persistent memory, deployed endpoint
Adversarial cases, CI gate
Least privilege, no agent sprawl
You are assessed on what you ship, not on attendance. The final Sunday is Demo Day: your system running live in front of an invited industry panel, as your final assessment and a showcase of what you shipped.
At least six of seven merged on time.
Eval set includes adversarial cases and catches a real failure.
Red team findings addressed, identity and gateway configured.
System runs live in front of the panel.
Twenty seats and a screened entry, so nobody spends eight Sundays out of their depth.
Before enrolment is confirmed: call the Gemini API, force a strict JSON schema, handle one failure, push to a repo. Pass or fail.
Intermediate Python, Git and the command line, and you have consumed a REST API before. No prior LLM experience is needed.
Thirty two contact hours, under LKR 3,000 per hour. The cloud environment and every lab, dataset and mock API are included, so there is nothing further to buy.
Same fee, LKR 95,000 per seat, for a separate cohort of twenty. Curriculum announced soon. Enrol on the Microsoft Track
A full day elective, priced separately, open to cohort learners and to non cohort developers. Pricing confirmed on application.
Every session ends with something merged. The next Sunday opens by reviewing it on screen. Click any session to expand.
Where the line sits between a chatbot, a workflow and an agent. Model selection as engineering. Structured output as the foundation.
Tool design as API design. Debugging tool choice from logs. MCP local and remote. Grounding, chunking, hybrid retrieval, citation.
The hardest session. Deterministic paths versus model decided paths. Approval gates as architecture, not fallback. Callbacks and guardrails.
Sessions and state versus memory. Memory schema design and the cost of remembering too much. Deployment, rollback, budget caps.
What deserves to be an agent, what should stay code. Agent Cards and signed identity. Failure design for distributed agents. A2A versus MCP.
Eval sets that mean something. Trajectory versus final answer, because the right answer by the wrong route is still a bug. Drift in production.
Prompt injection, tool poisoning, data exfiltration, over permissioned agents. Least privilege for non human actors. CBSL, PDPA and regional framing.
Fix what the red team found. Model downgrade paths, caching, where token spend actually sits. The architecture document that outlives the code.
