AI Agents in Production
Beyond Chatbots: Building Reliable AI Systems
Hands-on workshop for engineers and technical founders. Design agents that survive contact with real users — retries, tool choice, evals, cost ceilings, and the failure modes nobody writes about.
- Duration
- 3 sessions × 90 min
- Mode
- Live online
- Audience
- CTOs · Engineers · Technical founders
- Schedule
- Next cohort opens soon · join the waitlist
01 · What you'll learn
Concrete outcomes by the end
- Pick the right agent topology for your problem (single-shot vs orchestrated vs swarm)
- Design tool schemas that minimize hallucination and latency
- Build an eval harness you can actually trust
- Set cost + latency ceilings without dropping output quality
- Operate agents in prod: retries, circuit breakers, graceful degradation
02 · Agenda
What we cover
Topology & design
Session 1Single-call vs orchestrated vs swarm. Where each pattern wins, where each one breaks.
Tool design
Session 1How to write a tool spec the model will actually pick correctly. Common schema pitfalls.
Evals + observability
Session 2Building a reproducible eval set, scoring rubrics, and what to log per turn.
Cost + latency control
Session 2Token budgets, model fallbacks, caching, and graceful degradation.
Production patterns
Session 3Retries, circuit breakers, fallback chains, and operating a 24×7 agent fleet.
03 · Who should attend
The right audience
04 · Prerequisites
Come prepared
- Comfortable with Python or TypeScript
- Have called an LLM API at least once
- Bring an agent you want to ship — we'll review live
05 · Speaker
Hosted by
Pankaj Kharkwal
Founder, Pankh AI
Pankaj builds production AI systems for businesses and runs Pankh AI. He has shipped agents, RAG pipelines, and observability stacks for companies that needed AI to actually work — not just demo.
06 · Outcomes
Why people attend
After this workshop you leave with a concrete artefact you built live and a playbook you can use the next week. Cohort chat stays open so you can ask follow-up questions while you ship.
07 · FAQ
Common questions
Which models / frameworks do you cover?+
Model-agnostic patterns. We demo with OpenAI/Anthropic/Azure OpenAI. Framework-agnostic — bring LangGraph, your own glue, or nothing.
Is there a code-along?+
Yes. Each session has a live build. You'll leave with a starter repo wired to a real eval harness.
More workshops