A PhD student in your field just submitted a systematic review. They used an AI pipeline to triage 800 abstracts in two hours, extract structured findings from each, and group them by year. You read 40 abstracts over the same weekend. What does that gap look like compounded over a career?
That's... a lot of papers I'm not reading. And a lot of synthesis I'm not doing because I don't have time.
This track closes that gap. You'll build agents that summarise, classify, extract, and search — using real AI models on real research text. Rate yourself on six AI skills before we start the first agent.
I've used ChatGPT for drafting. Is that what we're building here?
We're building structured, programmatic agents — not chat. The difference is reproducibility and control. Rate honestly where you are today.
Thirty days. One real project: an AI literature summariser. By Day 28 you'll have an agent that takes a research question, uses perplexity/sonar to pull the 5 most recent relevant papers, extracts structured findings from each, and returns a markdown mini-review grouped by year.
| Week | Focus | Research framing |
|---|---|---|
| 1 | First agent, system prompts, classification | "Summarise this abstract in two sentences for a lit review." |
| 2 | Structured output, Pydantic extraction | "Extract claim, authors, year, journal from this paragraph." |
| 3 | Pipelines, batch processing | "Classify 50 open-text survey responses into themes in one call." |
| 4 | Web search, tools, capstone | "Build the AI mini-review agent on a live research question." |
Every lesson runs real models against real research text. Tests check output shapes, not fixed strings.
A PhD student in your field just submitted a systematic review. They used an AI pipeline to triage 800 abstracts in two hours, extract structured findings from each, and group them by year. You read 40 abstracts over the same weekend. What does that gap look like compounded over a career?
That's... a lot of papers I'm not reading. And a lot of synthesis I'm not doing because I don't have time.
This track closes that gap. You'll build agents that summarise, classify, extract, and search — using real AI models on real research text. Rate yourself on six AI skills before we start the first agent.
I've used ChatGPT for drafting. Is that what we're building here?
We're building structured, programmatic agents — not chat. The difference is reproducibility and control. Rate honestly where you are today.
Thirty days. One real project: an AI literature summariser. By Day 28 you'll have an agent that takes a research question, uses perplexity/sonar to pull the 5 most recent relevant papers, extracts structured findings from each, and returns a markdown mini-review grouped by year.
| Week | Focus | Research framing |
|---|---|---|
| 1 | First agent, system prompts, classification | "Summarise this abstract in two sentences for a lit review." |
| 2 | Structured output, Pydantic extraction | "Extract claim, authors, year, journal from this paragraph." |
| 3 | Pipelines, batch processing | "Classify 50 open-text survey responses into themes in one call." |
| 4 | Web search, tools, capstone | "Build the AI mini-review agent on a live research question." |
Every lesson runs real models against real research text. Tests check output shapes, not fixed strings.
Create a free account to get started. Paid plans unlock all tracks.