For students working on projects in our lab
Goal: create a lightweight project plan, decision record, and handover path that another authorized person can understand.
Prerequisites: project purpose, supervisor, expected duration, and known deliverables.
Execution context: planning in the supervisor-approved project repository or collaboration system. Do not put confidential data or credentials into a generic task board or AI prompt.
This is a lightweight guide to help you plan and run your project effectively without spending excessive time on management.
The aim: stay organized, deliver results, and learn along the way.
1. Define Goals (SMART Principle)#
Every project should start with clear objectives.
Use the SMART framework:
- Specific → clearly describe what should be achieved
- Measurable → define how progress or success will be checked
- Achievable → realistic with the given time and resources
- Relevant → connected to the project purpose and learning goals
- Time-bound → linked to a clear deadline
👉 Example: “By month 2, have a prototype that can operate continuously for 1 hour.”
2. Milestones & Timeline#
Break the project into 3–6 major milestones with deadlines.
| Milestone | Deadline | Deliverable |
|---|---|---|
| Literature review complete | Week 3 | Short summary |
| First prototype | Week 6 | Testable setup |
| Data collection finished | Week 10 | Dataset |
| Final presentation/report | Week 14 | Demo + written report |
3. Roles & Responsibilities#
Define who does what early on. One person can take multiple roles.
| Role | Person | Notes |
|---|---|---|
| Project coordinator | … | Organizes meetings, keeps overview |
| Technical lead(s) | … | Responsible for methods/design/build |
| Documentation lead | … | Notes, reports, presentations |
| Communication lead | … | Point of contact with supervisor |
4. Weekly Meetings & Tracking#
Hold short weekly check-ins (20–30 min):
- What was done last week?
- What’s next?
- Any blockers or help needed?
Meeting notes template (minimal):
- Key decisions
- Action items (who, what, by when)
👉 Keep a shared task list (Trello, Notion, Excel, whiteboard — whatever works).
For AI-native projects, add one lightweight line to the weekly notes:
- AI/agent use: what was delegated, what was accepted, what was rejected, and which checks were run.
5. Simple Project Planning#
- Split the work into chunks (e.g. background research, design, experiments, analysis, documentation).
- Place them on the timeline (e.g. in Excel or on a wall).
- Note dependencies (e.g. “analysis starts after data collection”).
6. Risks & Mitigation#
Think about potential obstacles early. Keep a short risk table (3–5 items).
| Risk | Likelihood | Impact | Mitigation |
|---|---|---|---|
| Equipment not available | Medium | High | Reserve in advance, plan alternatives |
| Data loss | Low | High | Regular backups |
| Team member unavailable | Medium | Medium | Share knowledge, cross-train |
Add AI-specific risks when relevant:
| Risk | Likelihood | Impact | Mitigation |
|---|---|---|---|
| Agent produces plausible but wrong code | Medium | High | Require tests, smoke checks, and human diff review |
| Secret or unpublished data enters a prompt | Low | High | Use approved tools only; never paste tokens or sensitive data |
| GPU jobs waste shared resources | Medium | High | Start CPU-only; use tiny smoke tests; batch long runs |
7. Decision Records and AI Assistance#
For important technical choices, keep a short decision note:
Date:
Decision:
Options considered:
Why this option:
Verification:
AI assistance used:
Open questions:
This is enough to explain later why the project chose a dependency, model, dataset split, solver setting, or Euler resource pattern.
8. Key Takeaways#
- Start with clear SMART goals
- Define milestones and roles
- Meet weekly and track tasks
- Use a simple plan to organize the work
- Identify a few risks and prepare mitigation
- Record important decisions and AI assistance
➡️ These steps should be enough to keep your project structured and running smoothly, while leaving you time to focus on research and experimentation.
Use the Project Handover Lab to turn this plan into reproducible evidence for the next researcher.
Verification#
- The plan names an owner, milestones, deliverables, and near-term actions.
- Important technical decisions record evidence and unresolved risks.
- Code, data, access, and final results each have an identified handover owner.
- Weekly notes do not contain secrets or unnecessary confidential material.
Understand Before Accepting AI Output#
- I chose the priorities and deadlines rather than accepting invented ones.
- I verified every claimed result and dependency.
- The information owner approved any protected context given to the tool.
- I can explain which risks require supervisor action.
Common Failures And Safe Recovery#
- Plan grows into bureaucracy: retain only decisions, owners, milestones, risks, and actions that help delivery or handover.
- Agent invents progress: correct the record from actual commits, results, meeting decisions, and measurements.
- Ownership is unclear: stop moving or sharing project assets and ask the supervisor to name the responsible person.
Ask For Help When#
The scope, data owner, access owner, publication deadline, compute budget, or handover destination is unclear.
Verified: 2026-08-04. Review by: 2027-02-04. Owner: project supervisor.