Teaching
The Maven cohort
AI Problem Framing for AI Practitioners. This is for the practitioner who has built enough AI to be frustrated by it: the approach that turned out wrong, the results that underwhelmed, the sense that the next one could go better. The course runs four weeks. You bring a real project, and each week you run it through the frameworks in a live session where we work through it together.
The course lessons
- Mindset. Build the instinct to frame a problem from first principles, and see what a wrong frame actually costs.
- The GOATS Loop. Get a repeatable way to frame a project before you build it: the five GOATS steps, what you are actually predicting (the atomic unit), and how much to let the system act on its own (the autonomy spectrum).
- The Diagnosis. Learn to read the signals that tell you a frame is broken, so you catch it before it gets expensive.
- The Pivot. When a project stalls, make the call to persist, pivot, or stop, instead of drifting.
The next cohort runs September 2026, and early-bird pricing is open now. Sign up on Maven →
What students say
Students rate the course 4.9 on Maven. A few words from the people who took it:
No matter what your experience level is, this is a great overview for anyone working in the space of applied AI.
— Taylor Larkin, Senior Data Scientist, DataRobot
Rajiv brings rigor and relevance together, while avoiding the AI hype and focuses the course on helping you frame problems in a way that make them suitable for agents and then evaluating performance and also deciding when to discontinue projects. Strongly recommend.
— Vishal Sachdev, Professor, University of Illinois at Urbana-Champaign
This is a timely and valuable course worthy of your time and expense. The Domain Specific Checklists alone justify it because you can use them immediately in your work. I especially appreciated Lessons 3 & 4, which lie at the heart of the course. We are all becoming AI managers, which means our decision-making skills are our most valuable (and marketable) skills. This course hones those within the context of applied AI, which makes it particularly relevant for AI practitioners and decision-makers.
— Chad Harness, Head of Technical Enablement, Snorkel AI
Rajiv is an excellent instructor who's meticulous and thoughtful in his course design. His content is approachable from all skill levels, offering something for the newer entrants and seasoned pros alike. I found the AI Problem Framing series incredibly insightful for my own professional work. The sections on selecting the right tech and model type for the right problem (vs. treating an LLM like a magic wand to solve everything) were really helpful in making this a repeatable part of project work going forward.
— Jake Shaver, Chief of Staff
My attitude to AI projects has completely transformed as a result of this training. It stresses defining the proper problem rather than concentrating on models first, which is where most AI initiatives fall short. Particularly useful is the "Loop" architecture, which offers an organized and transparent method for defining issues, weighing trade-offs, and agreeing on success measures. This course is perfect for anyone who wishes to approach AI more strategically rather than only creating models.
— Doan Huynh Phuc Thinh, NAB
Exceptional content that I've never seen elsewhere. Content is really insightful.
— Ramdas Krishnakumar, AI Engineer, Panasonic
A very insightful course with practical takeaways. I'm looking forward to applying what I learned to help drive our AI strategy.
— Isabel George, Executive Director of Product Management, Medscape
About the instructor
Rajiv Shah is an AI Agentic Engineer at OpenHands. He has spent the last decade helping teams ship AI systems that hold up in production, across industries from finance to healthcare to enterprise SaaS. He teaches and writes about the parts of AI work that do not get covered in tutorials.
Other things to read
If this site is useful, the Learn Harness Engineering with OpenHands tutorial is the engineering-side companion: how to build, instrument, and modify the system around the model after the framing decisions are settled.