AI, Agents & NASA: An Evening with IBM ResearchWelcome to an evening with IBM Research where we discuss AI, Agents, NASA and everything in between.
An event hosted by **IBM Research** \- PyData Ireland is excited to be a community partner\.
This is an inaugural **Open Source Science Dublin** meetup, where cutting-edge scientific research meets open-source technology, AI, and cloud-native infrastructure.
Expect **technical deep dives**, **cross-disciplinary conversations**, and (of course) **pizza** 🍕.
For this first event, we’re exploring one of the most exciting frontiers today: **agentic systems for scientific discovery**, and how autonomous and semi-autonomous AI systems are transforming the way research is conducted, validated, and scaled.
You’ll see real-world systems in action, hear from both applied and research perspectives, and walk away with a clearer picture of how agents are moving from hype to scientific infrastructure.
## **Talks**
**Accelerated Knowledge Discovery at NASA: Building an Agentic AI Research Companion**
* **Speaker:** James Barry, Staff Research Scientist, IBM Research
* **Description:** A practical look at NASA’s AKD platform: a chat-driven frontend orchestrating a multi-agent backend (planner + literature/data/code search) to turn questions into traceable, end-to-end scientific research—faster, more systematic, and with human oversight.
**Production-Ready AI Agents: Containerization, Sandboxes, and the LangChain Stack**
* **Speaker:** Fabio Lorenzi, Staff Research Scientist, IBM Research
* **Description:** Building secure, scalable AI agents for industrial time series analysis using isolated code execution, LangGraph for agent orchestration, and FastAPI for production deployment—bridging foundation models and real-world maintenance workflows.
**Science in the Agentic Era: Structured Experimentation with *ado***
* **Speaker:** Michael Johnston, STSM, Discovery Systems; Manager, Next Generation Systems, IBM Research
* **Description:** A research-first framework for agent-assisted discovery: *ado* encodes the problem space and experimental plan as schemas, enabling agents to propose and refine studies while keeping every run transparent, reproducible, and scientifically auditable.