Problem statement A: Meeting Action Items Tracker (mini workspace)
Build a web app where I can:
paste a meeting transcript (text is enough)
get a list of action items (task + owner if possible + due date if found)
edit, add, and delete action items
mark items as done
See a simple history of the last 5 transcripts I processed
Make it your own: for example, add filters (open/done), tags, or a better editor.
Problem statement B: Workflow Builder Lite (small automation runner)
Build a web app where I can:
create a simple “workflow” with 2–4 steps(example steps: “clean text”, “summarize”,
extract key points”, “tag category”)
run the workflow on an input text
see the output of each step
see a simple run history (last 5 runs)
Make it your own: for example, you can add step types, templates, or a nicer run screen.
What to include
A simple home page with clear steps
A status page that shows health of backend, database, and llm connection.
Basic handling for empty/wrong input
A short README: how to run, what is done, what is not done
A short AI_NOTES.md: what you used AI for, and what you checked yourself. Which LLM and
provider does your app use and why.
Put your name and resume in [Link]
A PROMPTS_USED.md, with records of your prompts used for app developemnt. Don’t
include agent responses, api keys, etc.
Hosting requirement
Hosting is important for this role.
Keep the app live after you submit.
If you truly cannot host, you may submit without hosting only if:
the repo runs with one command using Docker, and
you explain why hosting was not possible
This will be scored lower than a hosted submission.
If your link is down during review, you get one chance to fix it.
Rules for Github repo
Do not put API keys or passwords in the code.
Use an .[Link] file for settings.
How we review
we will open the live link and try a few actions. Then we will read your GitHub repo. We look for:
it works and is easy to use
clean code
basic checks and testing
sensible use of AI (not blind copy-paste)