Most placement advice sounds like this: "Learn DSA. Do LeetCode. Build a CRUD app."
That advice isn't wrong. But it's also what every other candidate is doing.
Here's what's shifted in the last 18 months: recruiters — especially from product companies and startups — are now filtering candidates on AI tools, prompt engineering, and practical AI project experience. Not as a bonus. As a first-round filter.
The candidates who get shortlisted aren't always the ones with the highest CGPA. They're the ones who can demonstrate they've worked with AI practically and can talk about it with specificity.
This checklist covers how tech students can build that edge in 90 days — alongside a regular semester, without quitting LeetCode.
Before You Start: One Honest Expectation
Prompt engineering is not a magic skill. You won't "master AI" in 90 days. What you can do is go from zero practical experience to one solid project, a clear interview narrative, and enough hands-on time to speak with confidence about how you work with these tools.
That's enough to stand out. Most of your competition has done nothing.
Month 1 — Learn to Use AI Tools, Not Just Chat With Them
Most students use ChatGPT or Claude for quick answers and call it AI experience. Recruiters see through this in about 30 seconds.
Real AI literacy means understanding why a prompt works — not just that it produced something useful.
Week 1–2: Core Concepts Worth Actually Understanding
- The difference between zero-shot, few-shot, and chain-of-thought prompting. Don't just read about them — test each one on the same problem and compare outputs side by side.
- What a system prompt is and why it changes model behaviour entirely. Anthropic and OpenAI both publish clear documentation on this for free.
- Pick one tool and go deep — Claude, ChatGPT, or Gemini. Depth beats breadth at this stage.
If you want solid technical grounding before experimenting, this practical guide to prompt engineering for developers covers the mechanics without the hype.
Week 3–4: Three Exercises That Actually Build the Skill
Exercise 1: Take any assignment you're currently working on. Solve it yourself first. Then prompt an AI to solve it. Compare where the outputs diverge and why.
Exercise 2: Write a prompt that produces a bad output. Rewrite it three times until the output is genuinely useful. Save all four versions. This is your first portfolio artifact — it shows iteration, which is exactly what interviewers want to see.
Exercise 3: Use an AI tool to generate code for something you already understand — a sorting algorithm, a simple API call. Then read every line and explain it out loud as if you're in a viva.
The third exercise matters because interviews will test whether you understand AI-assisted code, not just whether you can generate it.
Month 2 — Build One Real Project (Small Is Fine, Real Is Non-Negotiable)
This is where most students stall. They want to build something impressive before they've built anything. That's the wrong order.
Build something small that solves a real problem. The best source of real problems? Your own campus.
Project Ideas That Work Well
Exam prep assistant
Feed your university's previous year question papers into an AI pipeline. Let it generate practice questions, topic summaries, and explanations on demand. Every student in your batch is a potential user — that's a real user base, not a hypothetical one.
Placement FAQ bot
Before every campus recruitment drive, students ask the same 20 questions. Build a simple bot that answers them using your college's placement cell data. Your placement officer will likely appreciate being looped in — and that's a real stakeholder, which looks strong in any interview.
Assignment brief analyser
A tool that takes a vague assignment brief and returns a structured breakdown — key requirements, suggested approach, likely evaluation criteria. Genuinely useful, genuinely buildable in a weekend.
Timetable conflict checker
Use AI to parse timetable data and flag clashes or suggest study blocks. Boring problem, useful solution, great talking point in interviews.
None of these require cloud infrastructure, a team, or a paid API tier. A working prototype with real feedback from even five classmates is worth more in an interview than an ambitious project nobody actually used.
What to Document as You Build
This part is as important as the building itself:
- Every prompt you wrote that didn't work, and why you think it failed
- The iteration that finally worked
- One measurable outcome — even "8 batchmates used it during exam week" counts
This documentation becomes your project story in interviews. It's the difference between "I built a chatbot" and "Here's what I tried, what broke, and what I changed."
Month 3 — Package It for Placements
The skill exists. The project exists. Now it needs to be presentable in a way a recruiter understands in 30 seconds.
GitHub README — The Most Ignored Placement Asset
Your README should answer three questions without requiring the reader to dig:
- What does this project do, in one sentence?
- What AI or prompt engineering decisions did you make, and why?
- What surprised you — what didn't work the way you expected?
The third question is what proves you actually built it rather than just describing a concept you read about.
LinkedIn Education Section
Add one specific line under your degree: "Built [project name] using prompt engineering and [tool] — used by [X students / solved Y problem]."
Specific beats vague in every context. "Built an AI project" tells a recruiter nothing. "Built an exam prep tool used by 40 students before end-semester exams" tells them everything.
What to Say in the Interview
Recruiters asking about AI skills are probing two things: whether you've used it practically, or whether you understand its limits. Prepare for both.
For practical use: Have your project open on your laptop. Walk them through one prompt iteration — before and after. Show the thinking, not just the output.
For limitations: Be direct. "It hallucinated here, so I added a validation step" is a stronger answer than pretending the tool was perfect. Showing you understand failure modes signals genuine experience — and that's rare among fresh graduates.
Which Companies Are Actually Looking for This in 2026
From placement patterns across engineering colleges, the consistent demand is coming from:
- Product startups (Series A and above) — they want students who can build AI-assisted features without hand-holding
- IT services firms building AI practice teams — upskilling fast, want students who don't need to start from zero
- EdTech companies — particularly relevant if your project touches education in any way
- GCCs (Global Capability Centres) — hiring has significantly picked up and AI literacy is increasingly a differentiator
Worth noting: this isn't limited to CS or IT roles. Design-facing tracks are shifting too. Fields like product design, UX, and specialised creative roles are incorporating AI tools into everyday workflows — and entrance into those tracks is getting more competitive. If you're mapping options beyond traditional engineering placements, Shapeverse tracks how design entrance exams like UCEED, NIFT, and NID are evolving — useful context for anyone considering design as a parallel path.
The University Side of This Equation
One thing worth understanding as you build: universities themselves are now being found and evaluated through AI search — not just Google.
How a university appears when a student asks ChatGPT "which college should I join for computer science in Gujarat" is becoming as important to institutions as their traditional search ranking. If you're building anything at the intersection of students and institutions — a placement tool, an admissions assistant, a course recommender — understanding this context makes you a sharper builder.
This piece on how AI is changing the way students discover universities breaks down the mechanics clearly. Written for university teams, but the underlying logic is directly useful for anyone building in this space.
The Full 90-Day Checklist
Month 1
- [ ] Test zero-shot, few-shot, and chain-of-thought prompting on the same problem — compare outputs
- [ ] Pick one AI tool and spend the month going deep on it
- [ ] Complete the three exercises above
- [ ] Save your prompt iterations as portfolio material
Month 2
- [ ] Identify one real campus problem and build a working prototype
- [ ] Document every prompt that failed and why
- [ ] Get real feedback from at least 3–5 actual users from your batch
- [ ] Note one measurable outcome, however small
Month 3
- [ ] Write a GitHub README that answers the three questions above
- [ ] Add one specific, measurable line to your LinkedIn education section
- [ ] Prepare your project walkthrough — have it open, know the story cold
- [ ] Prepare your limitations answer — be honest about where AI failed you
Ninety days is enough to go from nothing to something concrete. The students who get shortlisted in campus placements aren't always the most technically advanced — they're often the ones who started earlier and can talk about what they built with honesty and specificity.
Start with Month 1, Week 1. Pick one exercise. Do it today.
Working on your first AI project and hit a wall? Drop your question in the comments.
Top comments (0)