What This Is Not
This is not a request to build a fully polished trading application by next week. It is not a competition to produce the most impressive-looking dashboard. It is not an invitation to skip the Scrum process in favour of solo coding.
This is a Scrum engineering sprint. Your team decides what to automate first. Your Product Owner defines the sprint goal. Your Scrum Master protects the process. Your team builds, tests, and presents a working increment — however small — by next Monday.
Answer these four questions in your sprint planning session. The answers become your Sprint 4 backlog.
The Scrum Lens — Why This Is Still an Agile Exercise
You are not pivoting away from Scrum. You are applying Scrum to a new type of work. The sprint rhythm stays identical. The ceremonies stay identical. The only thing that changes is the nature of the increment — it is now software, not a prediction deck.
What is the sprint goal? Which one automation will you ship? What is your Definition of Done for a software increment?
What did I build yesterday? What will I build today? What is blocking the pipeline? Fifteen minutes maximum.
Demo the working software. Does it run? Does it produce the expected output? Show it live — not a screenshot.
What slowed down the build? What would you automate differently? What enters next sprint's backlog?
Three Paths Forward — All Valid, All Assessed on Execution Quality
These are not team rankings. They are sprint goals. A strong team at Level 2 that delivers a clean, working, well-documented increment will score higher than a weak team that overclaims Level 4. Choose the level that stretches your team without breaking it.
Automate one data collection step. Produce a structured output file. Commit it on a schedule.
DoD: Script runs. Output file appears in repo. Team can explain every line of code.
Multi-step automated pipeline with at least one LLM API call and a structured prediction output.
DoD: Pipeline runs end-to-end with one command. Output is inspectable. LLM call is logged.
Full intelligent pipeline with UI, multi-LLM synthesis, automated delta scoring, and prescriptive output.
DoD: System produces a complete weekly intelligence report without manual data entry.
What Every Team Must Submit — Present Monday 15 June
The prediction and release tag remain mandatory for all teams. The software increment is the new addition. The increment does not have to be large — it has to be real, working, and explained.
| Deliverable | Who | Format | Mandatory? |
|---|---|---|---|
| W4 Market Prediction SPX, NDX, IWM + at least 3 S&P sectors |
Full team | Prediction file in repo | Required — all teams |
| GitHub Release Tag vW24 Seals the prediction before market open |
R9 GitHub Lead | vW24 tag on repo |
Required — all teams |
| Software Increment At least one automated step from your chosen tier |
Dev Team | Code in repo + README | Required — all teams |
| Sprint Demo Run the software live in Monday presentation — not a screenshot |
R9 + Dev Team | Live demo, Monday 15 June | Required — all teams |
| W3 Delta Report What was your W3 prediction accuracy? Direction correct? Error size? |
R6 + R10 | Delta file or table in repo | Required — Level 2+ |
| Architecture Decision One paragraph: what did you choose to automate first and why? |
Product Owner | DECISION.md in repo | Required — Level 2+ |
| Sector Coverage (3+ sectors) Extend prediction to include sector signals — not just indices |
R3, R4, R5 | Sector analysis in prediction file | Expected — Level 3+ |
| LLM API Integration At least one model called programmatically, response logged |
R8 LLM Operator | Script + logged response in repo | Expected — Level 3+ |
| Prescriptive Output System-generated Rx based on W3 delta |
R7 + R10 | Rx file or dashboard output | Distinction — Level 4 |
Five Things That Will Not Earn Marks
If the data is hardcoded or the charts are static images, it is not a software increment. It is a mockup. Mockups do not fulfil the sprint goal.
Every team member must be able to explain the code in the Monday demo. If the Scrum Master cannot describe what the LLM call does, the team has a knowledge silo problem — not a software success.
The prediction is still the sprint increment. The software is a tool that helps produce better predictions. Teams that submit a demo but no prediction have missed the point entirely.
AI-assisted coding is encouraged. AI-generated code that the team cannot explain is a risk. During the demo, Professor Dr. Tan may ask any team member to explain any line. Prepare accordingly.
Be honest in your sprint planning about what your team can deliver. A modest, complete, working Level 2 increment delivered on time scores higher than an ambitious Level 4 that does not run.
Questions to Answer Before You Write a Single Line of Code
The last question is the most important. If your team cannot answer it, reconsider what you are building.