agents4science

Agentic Scientific Discovery Platforms

The following draft curriculum outlines topics to be covered and potential readings.

---

Week 1

Mon Sept 29 — Lecture 1: What is an SDP?

Introduces the concept of Scientific Discovery Platforms (SDPs): AI-native systems that connect reasoning models with scientific resources. We’ll explore motivating case studies (wildfire hazard, antimicrobials, climate modeling) and outline the challenges of integrating AI into rigorous science.

Readings:

• Exploring Large Language Model based Intelligent Agents: Definitions, Methods, and Prospects, Cheng et al. (Arxiv, 2024).
• Artificial intelligence and illusions of understanding in scientific research, Messeri & Crockett (Nature, 2023).
• The Shift from Models to Compound AI Systems – The Berkeley Artificial Intelligence Research Blog.

Wed Oct 1 — Lecture 2: Frontiers of Language Models

Surveys frontier reasoning models: general-purpose LLMs (GPT, Claude), domain-specific foundation models (materials, bio, weather), and hybrids. Covers techniques for eliciting better reasoning: prompting, chain-of-thought, retrieval-augmented generation (RAG), fine-tuning, and tool-augmented reasoning.

Readings:

• Chain-of-Thought Prompting Elicits Reasoning in Large Language Models.
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning.
• ReAct: Synergizing Reasoning and Acting in Language Models.

Assignment A1: Implement a ReACT style agent.

---

Week 2

Mon Oct 6 — Lecture 3: Systems for Agents

Discusses architectures and frameworks for building multi-agent systems, with emphasis on inter-agent communication, orchestration, and lifecycle management.

Readings:

• AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation.
• LangGraph.
• AIOS: LLM Agent Operating System.

Wed Oct 8 — Lecture 4: Retrieval Augmented Generation (RAG) and Vector Databases

Covers how to augment reasoning models with external knowledge bases, vector search, and hybrid retrieval methods.

Readings:

• Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks.
• The FAISS library.

Assignment A2: Hybrid retrieval.

---

Week 3

Mon Oct 13 — Lecture 5: Tool Calling

Introduces methods for invoking external tools from reasoning models. Focus on model context protocol (MCP), schema design, and execution management.

Readings:

• Introduction - Model Context Protocol.

Wed Oct 15 — Lecture 6: HPC Systems and Self Driving Labs

How SDPs connect to HPC workflows and experimental labs. Covers distributed coordination, robotics, and federated agents.

Readings:

• Self-Driving Laboratories for Chemistry and Materials Science, Chemical Reviews.
• Empowering Scientific Workflows with Federated Agents.

Assignment A3: Implement Distributed Battleship (and/or Implement MCP toolbox).

---

Week 4

Mon Oct 20 — Lecture 7: Human–AI Workflows

Explores how scientists and agents collaborate: trust boundaries, interaction design, and debugging.

Readings:

• Guidelines for Human-AI Interaction, Amershi et al. (CHI, 2019).
• Interactive Debugging and Steering of Multi-Agent AI Systems (CHI, 2025).

Wed Oct 22 — Lecture 8: Benchmarking and Evaluation

Frameworks for assessing agents and SDPs: robustness, validity, and relevance.

Readings:

• MLE-bench: Evaluating Machine Learning Agents on Machine Learning Engineering.
• AI Agents That Matter.
• EAIRA: Establishing a Methodology for Evaluating AI Models as Scientific Research Assistants.

---

Week 5

Mon Oct 27 — Lecture 9: Failures and Safety

Examines why multi-agent systems fail and methods for safety and guardrails.

Readings:

• Why Do Multi-Agent LLM Systems Fail?.
• AGrail: A Lifelong Agent Guardrail with Effective and Adaptive Safety Detection.
• Improve accuracy by adding Automated Reasoning checks in Amazon Bedrock Guardrails.

Assignment A4: Implement evaluation harness.

Wed Oct 29 — Lecture 10: Case Studies

Case studies of SDPs in biology and materials.

Readings:

• The Virtual Lab of AI agents designs new SARS-CoV-2 nanobodies, Nature.
• The AI Scientist-v2: Workshop-Level Automated Scientific Discovery via Agentic Tree Search.

---

Week 6

Mon Nov 3 — Lecture 11: Novelty and Plagiarism

Explores originality, credit, and the risks of plagiarism in AI-generated science.

Readings:

• Can LLMs Generate Novel Research Ideas? A Large-Scale Human Study with 100+ NLP Researchers.
• All That Glitters is Not Novel: Plagiarism in AI Generated Research.

Assignment A5: Capstone project planning (novel contributions).

Wed Nov 5 — Lecture 12: Building Agents and Workflows

Pipelines, workflow composition, and self-improving systems.

Readings:

• AFlow: Automating Agentic Workflow Generation.
• DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelines.

Assignment A6: Generating HPC workflows.

---

Week 7

Mon Nov 10 — Lecture 13: Finetuning

Covers approaches to adapt agents with reinforcement learning and real-world training.

Readings:

• OpenPipe/ART: Agent Reinforcement Trainer.
• Agent Lightning: Train ANY AI Agents with Reinforcement Learning.

Wed Nov 12 — Lecture 14: Responsible SDPs

Discusses ethical and policy dimensions: dual-use concerns, bias, carbon footprint, open science vs IP.

Suggested Readings:

• Capabilities and risks from frontier AI (AI Safety Summit, 2024).
• UNESCO AI Ethics framework.
• Carbon-aware computing literature.

---

Week 8

Mon Nov 17 — Lecture 15: Scaling SDPs [SC week]

Strategies for scaling: distributed compute, HPC, cloud-native orchestration. Covers resilience, scheduling, and cost/energy considerations.

Suggested Readings:

• KubeFlow or Ray documentation.
• DOE report on AI for Science (2020).

Wed Nov 19 — Lecture 16: Automation in Practice [SC week]

Demonstration of automation pipelines with monitoring, logging, and adaptive workflows. Emphasis on debugging and error recovery.

Suggested Readings:

• MLflow for experiment tracking.
• Globus Flow for automation.

---

No class week of Nov 24 – Thanksgiving

---

Week 9

Mon Dec 1 — Lecture 17: Frontiers of SDPs

Explores frontiers: multi-agent collaboration, embodied co-scientists, integration with digital twins. Students speculate on SDPs in 2030.

Readings:

• The AI Scientist-v2: Workshop-Level Automated Scientific Discovery via Agentic Tree Search.
• Digital twin literature (manufacturing & climate).

Wed Dec 3 — Lecture 18: Capstone Prep + Peer Review

Students present draft capstone plans, receive structured peer critique, and refine. Instructor provides guidance on scope, deliverables, and evaluation.

Suggested Readings:

• Project management frameworks (Agile for research).
• Sample capstone projects from ML/AI courses.

---

Final Week

• Mon Dec 8 — No Class
• Wed Dec 10 — Final Class Meeting: Capstone Presentations

---

This site is open source. Improve this page.