Foresight Workshop: Bootstrapping Research Agents — Hands-On for Scientists
Bring your current research project, and together we will build autonomous AI agents assisting you in the work.
Important: this workshop is meant for researchers, especially in the field of life-sciences, whose work might benefit from AI automation. If you are conducting high-impact research, we want to amplify your possibilities as a part of our mission. If you are part of the extended Foresight Institute family (Æthos, Effective Altruism, etc.), you are still very welcome, so that together we can continue this workshop as a cyclic event. Otherwise, consider booking a workshop directly with Xemantic
Agenda
10:00 doors open
11:00–13:00 introduction to Agentic AI (part 1)
13:00 lunch
14:00–15:00 introduction to Agentic AI (part 2)
15:00 building together till late night :)
We will start with the introduction to Agentic AI and harness engineering, to build common understanding and bridge any knowledge gaps. We start from first principles (what is a context window, really?) and arrive at agents that read papers, query databases, run analyses, and report back. You'll understand why each piece works, which means you'll be able to debug, extend, and adapt them well after the workshop ends. These topics are evolving faster than academic curricula can follow, and require a cognitive-science framing as much as technical skills.
Prerequisite: minimal programming experience (e.g. Python).
Following workshops
This is the first in the series of events organized by the Berlin AI Node of the Foresight Institute and Xemantic, stemming from Kazik Pogoda's experience of teaching Agentic AI, and mentoring Agentic AI track during Berlin Bio × AI Hackathon. You can find more information here:
https://xemantic.com/ai/workshops/
What you'll leave with
A working prototype agent doing something useful in your domain
A clear mental model of the agent stack: context windows, system vs. user prompts, tool calling, structured output, memory, control flow
A documented Python example repository (with Kotlin and TypeScript mirrors for the curious) you can keep building from
The vocabulary to read agent papers and engineering posts without getting lost
What to bring
A laptop (8GB RAM minimum, Python 3.10+ installed)
We will try to utilize Foresight's local compute, otherwise Anthropic API key with at least $5 of credit might be useful
A research question, dataset, or workflow you'd like to make tractable with an agent — bring your real problem, not a toy one
Note: there are several external screens ready to be used at Foresight Institute