AI Research Assistant

Researchers need synthesis across dozens of sources. This capstone demonstrates grounded generation, citation enforcement, and academic tool integration.

A PhD student who highlights every claim with a page number — if they can't find the source, they say 'I don't know.'

Scenario

Question: 'Compare RAG eval metrics 2024–2025.' Agent retrieves 14 chunks from 9 papers, outputs table (RAGAS faithfulness, DeepEval, human eval cost) each row linked to DOI, exports references.bib.

Solution

In Capstone Projects, apply AI Research Assistant to this scenario: Question: 'Compare RAG eval metrics 2024–2025. Identify the inputs, run the technique, validate the output, and note one thing you would monitor in production.

Do this before moving to the next module — reading alone is not enough.

Open the Code Walkthrough below and run it locally. Change one parameter related to AI Research Assistant (e.g. model, temperature, top_k, or tool name), observe the difference in output, and write 2–3 sentences explaining what changed.

Highlighted lines show where AI Research Assistant happens in the code.

1# AI Research Assistant — minimal example2from openai import OpenAI3
4client = OpenAI()  # create API client5
6# Ask the model to explain this topic7response = client.chat.completions.create(  # core API call for AI Research Assistant8    model="gpt-4o-mini",9    messages=[10        {"role": "system", "content": "You explain ai research assistant clearly."},11        {"role": "user", "content": f"What is ai research assistant?"},12    ],13    temperature=0,14)15print(response.choices[0].message.content)  # show output for debugging

• Letting LLM cite from training memory
• Fixed-size chunks on academic PDFs
• No reranker on retrieval

Quick recap — the most important points from this module.