Is it free with membership?

Yes with Premium ($299/yr): every course is included. General members ($99/yr) pay half, $24.50 instead of $49.

What do I need before I start?

Basic understanding of Python; Familiarity with LLMs (OpenAI / LangChain basics); Understanding of APIs and system architecture.

Can my team take it together?

Yes. Institutional membership covers every seat under one agreement, and we run private cohorts for teams. Talk to us through the corporate training page.

Intermediate

Debugging Non-Deterministic Agentic Systems

Modern AI agents fail silently. This course teaches you how to instrument, trace, and monitor agentic systems using distributed telemetry, structured tracing, and AI-native evaluation tools. Move beyond logs and build observable, debuggabl…

5+
enrolled: 3 hrs
of instruction: 10
lectures: 4
modules

Taught by Abhishek Kumar

Debugging Non-Deterministic Agentic Systems

What's inside

Live sessions: 3 · 100 min
Decks & handouts: 6
Readings: 1

01 · What you'll be able to do

Five things you can ship afterwards.

Why traditional debugging fails

Why traditional debugging fails for agentic systems.

How to instrument LLM

How to instrument LLM agents with structured telemetry.

Designing trace pipelines

Designing trace pipelines using OpenTelemetry & Prometheus.

Debugging silent failures like

Debugging silent failures like hallucinations and infinite replans.

Using LangSmith

Using LangSmith for live tracing, evaluation, and reasoning inspection.

02 · Curriculum

What's in the course.

4 modules · 10 lectures · 100 min of sessions

01 · 42m02 · 37m03 · 21m

Each block is one module, sized by session length.

Why Logs Fail When Agents Start Thinking42m
Why Logs Fail When Agents Start Thinking - Slide Deck (PDF)Deck
Autogen Demo CodePDF

03 · Instructors

Taught by the people who built it.

Abhishek Kumar

Instructor · 14 courses on ADaSci

All courses by Abhishek

04 · Is it for you?

Who it's for, and what to bring.

Built for

AI Engineers building LLM applications Backend / Platform
Engineers moving into AI systems
MLOps / Observability Engineers
Technical Architects designing agentic workflows
Developers deploying RAG or multi-agent systems

Bring with you

Basic understanding of Python
Familiarity with LLMs (OpenAI / LangChain basics)
Understanding of APIs and system architecture
Interest in AI engineering or production AI systems

Next step · Prove it

This course prepares you for the CAASA credential.

The Certified Agentic AI System Architect (CAASA) exam tests the skills this course teaches. Members sit it at a discount.

See CAASA

05 · FAQ

Before you enrol.

Each of the 4 modules is a recorded live session, with slide deck and handouts, readings from that session. Watch at your own pace.

$49.00

Related credentials

Certified Agentic AI System Architect (CAASA)

The industry-recognized agentic AI certification for engineers, architects and AI leaders. Master multi-agent systems, orchestration, governance and production deployment in 30 hours.