Claude, your academic assistant
You've already tried ChatGPT — and it made up a source or a number that didn't exist. You've tried NotebookLM — but you copied and pasted so much that you still need 4 hours a day to finish what you had to do. You gave it a whole project — but you had to explain so many details that it was easier to do it by hand.
The problem isn't you. It's the way you've been using it so far — without the discipline over sources that your academic work demands. This workshop starts right there: Claude works on your actual files, not from memory, cites where each piece of information comes from, and teaches you your work rules once — you don't repeat them at every prompt.
Reserve a seat for me →You work on your files — not on textbook examples. You tell him what you want, he works on your real files, you see every step — and where every piece of information comes from — and you decide.
How it works
Concept immediately followed by applied exercise, on your laptop, on your examples.
| When | What we do |
|---|---|
| Tuesday, September 8 · 6:00 PM–9:00 PM | Session 1 · Fundamentals + literature and data |
| Thursday, September 10 · 6:00 PM–9:00 PM | Session 2 · Project management + skills and automation |
| Thursday, September 17 · 6:00 PM–7:30 PM | Office hours · on your work |
Session 1 · Fundamentals + literature and data
- What makes Claude Code different from a chat and how it works in your folder
- the first task on real sources (e.g. research policies), citing each conclusion
- how to write a rules file (CLAUDE.md) that it reads automatically
- literature synthesis on a corpus of works
- data analysis: code run on a CSV + honest graph
Session 2 · Project management + skills and automation
- organizing a legacy project folder (minutes, emails, notes, photos, timesheets)
- administrative reports and documents: activity report, budget execution, time sheets, requirement reports, specifications
- how to find and install a skill, then how to build your own (e.g. activity report skill)
- critical self-evaluation of one's own work with a panel of sub-agents
- Claude's connection to Semantic Scholar, PubMed, Consensus, Elicit, SciSpace, NotebookLM — no technical jargon
- routines: a task that runs on schedule; and where the automation stops
Office hours, September 17
1h30 on your own files, on the real frictions that have arisen in the meantime.
Each session is recorded and remains available after the workshop.
The problem we are solving
You rarely read just one paper: you have a stack, and you have to see where they contradict each other and what's missing. You have data that you have to gather correctly, not "from memory." You have your own paper, which you should look at critically before submitting.
And, in parallel, the project bureaucracy is not waiting: an inherited folder with meaningless file names, an activity report with a deadline to the funder, budget execution to explain, timesheets to fine-tune, reports of necessity and specifications to write. This is where the hours are spent — not on the actual research.
The workshop is for:
- researchers and university teachers
- doctoral and postdoctoral students
- project administrative staff (grant management, reporting)
- technology transfer specialists and project offices
- anyone who works with literature, data, reports, and funder deadlines
It's designed for academic research and project management — not general office work. (If you do PM, HR, or content without a research component, there's another workshop that's more appropriate, below under "Who it's not for".)
You don't need programming experience. You need real work with files — papers and datasets, but also project folders, reports, budgets, timesheets, purchases. (And yes: even if you're doing "just research," you still end up doing those things too.)
What will you be able to do in the end?
Each skill is practiced on a real case, on files — not on slides.
Literature synthesis on an entire corpus.
From a stack of papers — comparative table, cross-cutting themes, contradictions, limits and open questions, with each statement anchored in the source. What is not in the text goes to "to be verified", it is not invented.
Reproducible data analysis.
Claude runs code on your file (CSV, table), aggregates the numbers, and outputs an honest graph — with steps you can redo at any time.
Organizing a chaotic project folder.
You transform a legacy folder into a clear structure: what each file is (including pictures read as images), what actions and deadlines it contains, what contradictions exist — flagged, not smoothed over.
Project reports and administrative documents.
Activity report in the financier's format, budget execution, time sheets, requirement reports, specifications — starting from your real files, with figures taken from the source.
A critical look at one's own work, with subagents.
A panel of sub-agents, each with a different lens, reviews your work before submission: weaknesses, questions for the author, score. We legitimize your work before you send it anywhere.
Skills, connected tools, and routines.
You install a ready-made skill and build your own (e.g. the one that makes your activity report). You have Claude work with the tools you already use for literature, instead of jumping from one to another. Without digging through technical settings: you tell it what you want, in Romanian.
Claude on autopilot.
You set it to run overnight or at a fixed time. It's waiting for you in the morning with the day's agenda ready, reports already organized in folders, and drafts of responses to emails ready to be written — you just read them and press send.
At the end, you leave with a folder of examples, at least one skill of your own, and a way of working that you can immediately apply — both to research and to managing your projects. You can make up for the hours you wasted with copy-paste here.
Who teaches you?
I am Madalina Cocea. I have been working in scientific communication for 20 years — currently at the Technology Transfer Center of the "Alexandru Ioan Cuza" University in Iași. I have trained over 2,000 researchers from over 10 universities: how to communicate what they do, how to structure their work, and how to manage communication on complex projects.
I know both research and project management inside out — the literature, the data, the reporting to the funder, the deadlines. I use Claude daily in my own work, with the discipline on sources that academia demands. This workshop is built from what works in practice, not from theory.
Your reservations — and why they shouldn't be
Any serious researcher has the same reservations. You're right. We'll put them on the table.
„"How do I know he's not hallucinating?"”
If you send a wrong number or a made-up quote to a funder or a journal, it's your responsibility, not the AI's — and you already know that. That's why I don't let it work from memory. At the workshop, Claude reads the source you give him and runs code on your data. Every statement is anchored to where it came from; what it doesn't find in the source is marked "to be verified," not invented.
„"How do I check what he wrote without taking on an extra job?"”
Verification is built into the flow, not pasted on at the end: every conclusion has its source next to it, every number comes from code you can run again. You don't redo the work — you follow the path.
„"Aren't you stealing my ideas?"”
The methodology, thinking, and decisions remain yours — Claude does the reading and drafting, you keep the thread. And you control what you expose: at the workshop we establish exactly what is okay to show when it comes to unpublished or sensitive data.
„"How do I understand what it does, if the mechanism is complicated?"”
You don't work in a black box. You write to it in Romanian what you want, it works step by step in front of you — you see what it reads, what code it runs, what it writes — and you approve each step.
The entire workshop stands on a single line of integrity, the one required by funders and journals (COPE, ICMJE, Nature, JAMA, the European framework): AI as a tool, not as an author; transparency in declaration; confidentiality of sensitive data; the review for a journal remains human, as does the final decision. Check everything — both the citations and the figures.
What's included
- 2 live sessions of 3 hours each, applied, on Zoom
- Written guide + setup video, sent before the workshop
- Recording of each session, available after the workshop
- A folder with scenarios and real datasets
- Ready-to-install skills for research (literature synthesis, data profiling, review, activity report)
- At least one own skill, built during the workshop
- Certificate of participation
- Office hours follow-up on your own files
Places are limited to 25, so that I have time to respond to each one individually — at setup, and during office hours.
For whom it is not
If you want to fully automate article writing or bypass the integrity rules of journals and funders: this is not the workshop. Here we learn to use AI rigorously — with sources, verification, and human accountability.
And if you work in an office job (PM, HR, content, administrative) and want the same reflexes in your daily work, you have a dedicated course: Claude, your colleague at work.
What do I need?
- Own laptop (Windows or macOS), with rights to install a new application
- Account Claude with a paid plan that includes Claude Code — the most affordable, Claude Pro, starts at ~$20/month. (Claude Code, the tool we use, is only available on Anthropic's paid plans.)
- Stable internet connection and a Zoom account (free is enough)
Important: Claude's subscription is taken directly from claude.ai, not from other services that promise „access to Anthropic models.” It’s not the same thing and you won’t be able to work in the workshop with them.
Frequently Asked Questions
Do I need to know how to program?
No. You work through the Claude app — you write what you want to it, in Romanian, and it works on your files. The fact that it runs code on your data is an advantage (reproducible results), not something you have to write.
Do I have to give up my tools (Elicit, Consensus, SciSpace, NotebookLM)?
No. You keep using them for what they do well — search and discovery. Claude becomes the place where you gather what you find and transform it into a sourced, verifiable, and reproducible result: you connect it to your sources and literature indexes so that it starts from real works, not from memory.
What happens if I miss a session?
Each session is recorded, and you receive the recording after the workshop. Office hours are when we catch up on anything you missed in the meantime.
Registration
Online edition — September 8–17, 2026. 25 spots, so I can respond to everyone.
English (UK) 
Română