Build a WhatsApp bot for your college
Students ask the same questions every week: when do the exams start, is the campus closed tomorrow, is this topic even in the syllabus. Build a bot that answers them on WhatsApp from your college’s official documents only — with the source attached — and put it in front of real classmates.
Posted by The AI Internships
The work brief
- 01Collect 3+ official documents from your college or university: the academic calendar, your semester syllabus, and one rules document (attendance, exams, or grading). Record each one’s source link and the date it was published.
- 02Build the retrieval pipeline yourself so every answer names the document and section it came from, and says plainly when the documents do not cover the question.
- 03Handle the questions plain search cannot answer: “is the campus closed tomorrow”, “when is the first sessional test”, “how many classes can I miss”. These need today’s date resolved against the calendar and the thresholds read out of the rules — not just the nearest matching paragraph.
- 04Connect it to WhatsApp so anyone can message it, then get 10+ classmates who did not build it to use it for real.
- 05Log what they actually asked, score how the bot did, and fix the worst failure you find.
What you’ll produce5 deliverables
Submission standard
Submit your document set, a way for us to message the bot ourselves, the public repo, your scored test table, and the real question log from your classmates. Include your AI workflow: tools, best prompts, and what you changed from the AI’s output. Your repo’s commit history must show the pipeline coming together over more than one sitting, not a single upload.
The official documents you loaded, with source link and publication date for each
Name each document, where it is officially published, and the date or version on it. “A syllabus PDF someone shared” is not a source — we need the document a college office would stand behind.
Written responseRequiredHow to message your bot (number plus any join code or steps)
Write it so a stranger can message the bot in under a minute. If it runs on a sandbox, include the exact join code.
Short answerRequiredPublic GitHub repo with your code
Commit history should span multiple sessions (document loading, retrieval, date handling, WhatsApp wiring as separate dated commits), not one paste. A single dump commit fails the authenticity gate.
Public linkRequiredYour 12 scored test questions: 6 factual, 3 date-based, 3 traps — expected vs. what the bot actually said
RAG means retrieval-augmented generation. For each question give the expected answer, the bot’s actual reply with its citation, and pass or fail. The 3 traps must be questions your documents genuinely cannot answer.
Written responseRequiredThe worst failure your classmates exposed, and how you fixed it
Name one real failure from the log, explain why the bot got it wrong, and show what you changed.
Written responseRequired
You’ll complete these inside your private workspace.
What you must submit as proof2 proof items · Required
This brief requires evidence an AI can’t fabricate.
Log of 20+ real questions your classmates asked, unscripted
One header row, then one row per message: date, who asked (first name or initials), their question, what the bot answered, and whether it was right. Must come from 10+ different people who did not build the bot. Remove phone numbers.
CSV dataRequiredScreenshot of a real WhatsApp thread showing both a cited answer and a refusal
One WhatsApp screenshot from a real conversation showing the bot citing a document for one question and admitting it cannot answer another. The WhatsApp interface must be visible.
Image uploadRequired
Submissions without this evidence cannot be submitted.
Protect other people in your proof. Blur faces, names, phone numbers and email addresses before you upload, and refer to anyone you worked with by role or number ("Listener 1", "the stall owner"). Your proof is only ever used to check your work — it is never published, never appears on your certificate, and is never shown in your public portfolio.
How your work is evaluatedPass mark: 70/100
The passing benchmark is 70/100. Final score: Output 50% · AI process 30% · Evidence 20%. Required proof checks must also pass before certification.
Output criteria
Answers with real sources
34%Answers come from the official documents and name where they came from.
Handles dates and limits
33%Questions like “campus closed tomorrow?” and “how many classes can I miss?” resolve correctly against the calendar and the rules, not by keyword match.
Admits what it does not know
33%The 3 trap questions get an honest “not in the documents” instead of an invented answer.
Required proof checks
Used by real classmates
Certification is held for review if the log is invented, comes from fewer than 10 people, reads as scripted or self-written, or the bot cannot be messaged at the address given. The reflection must name a failure that log actually exposed.
Wired over time, not pasted
Fails if the repo is a single copy-paste commit, is private or empty, or the live bot does not match the code in the repo.
How we grade your AI usage30% of your score
Using AI is the point — it’s the skill this certificate proves. You’ll answer three short questions about how you used it: what you asked, what was wrong with its first answer, and what you changed. Specific, honest answers score high. “I pasted the brief and submitted the answer” scores near zero.