I Built a Career Advisor That Remembers You
The Problem Every Student Knows Too Well
It is the third week of March.
Priya, a third-year Computer Science student, opens an AI chat and types:
“Can you help me with my resume for a backend internship?”
The AI replies:
“Sure! What skills do you have?”
Priya sighs.
She answered this exact question last Tuesday. And the Tuesday before that.
Every time she opens a new session, she has to reintroduce herself — her skills, her projects, her goals. The AI is helpful, but it has the memory of a goldfish.
This is the invisible tax students pay when using AI tools for career preparation.
You spend the first few minutes of every session just getting the AI up to speed — and even then, the advice is generic, not personalized.
So I asked a simple question:
What if the AI actually remembered you?
Introducing the AI Internship & Career Advisor
I built an AI Career Advisor that doesn’t reset every time you open it.
Instead of treating each conversation as new, it uses persistent memory to track:
- Skills you’ve learned
- Projects you’ve built
- Internships you’ve applied to
- Interview outcomes
- Career goals
This is powered by a memory system called Hindsight, which allows AI agents to remember, recall, and improve over time.
Instead of acting like a chatbot, the system behaves like a mentor who has been following your journey for weeks.
A Day in Priya’s Life — With Memory
Session 1: Onboarding
Priya shares her profile:
- Python, SQL, learning React
- Built a finance dashboard
- Wants a backend internship
- Applied to Zepto
The system stores this in memory.
It will never ask again.
Session 2: Resume Help
Three days later, she returns:
“I have an interview at Meesho. Update my resume.”
No repeated questions.
The system already knows her background and responds:
“For a backend role, your finance dashboard project should be highlighted more — it shows API design and data handling.”
This is not generic advice. It is built from her actual history.
Session 3: A Rejection
Priya says:
“I got rejected from Meesho.”
The system remembers she was also rejected from Zepto earlier.
It responds:
“That’s two HR-round rejections. This might not be a technical issue — let’s practice your introduction.”
This kind of pattern recognition is only possible with memory.
Session 4: Skill Gap Analysis
She shares a job description from Razorpay.
The system compares it with her stored skills:
“You’re strong in Python and SQL, but missing API development experience. Here’s a 3-week plan to fix that.”
This is targeted, actionable guidance.
Session 5: Progress Tracking
Weeks later, Priya sees:
- 5 applications
- 2 interviews
- 1 in progress
The system reminds her:
“Your PhonePe application is 12 days old. Want to send a follow-up email?”
Nothing is forgotten.
How Memory Makes This Possible
This system runs on a simple loop:
Step 1 — Recall
Fetch user history from memory
Step 2 — Reason
Use that context to generate personalized responses
Step 3 — Store
Save new updates back into memory
This loop allows the system to evolve with the user.
This is powered by:
- Hindsight GitHub: https://github.com/vectorize-io/hindsight
- Documentation: https://hindsight.vectorize.io/
- Agent Memory: https://vectorize.io/features/agent-memory
How Memory is Stored (Conceptually)
Each interaction is stored in structured form:
{
"user": "Priya",
"skills": ["Python", "SQL"],
"projects": ["Finance Dashboard"],
"applications": ["Meesho", "Zepto"],
"outcomes": ["HR rejection"]
}
Before generating a response, the system retrieves this memory and uses it as context.
This is what makes responses personalized.
Before vs After Memory
Without memory:
“Tell me your skills again.”
With memory:
“Since you’ve worked with Python and built a finance dashboard, here’s how you can improve your backend profile…”
That difference is everything.
Core Features
1. Resume Feedback
Personalized suggestions based on real projects
2. Skill Gap Analysis
Identifies missing skills and creates learning paths
3. Internship Recommendations
Suggests relevant roles based on history
4. Progress Tracker
Tracks applications, interviews, and outcomes
Why This Matters
Most students don’t lack effort — they lack guidance.
AI tools today are powerful, but they forget everything.
By adding memory, we move from:
generic answers → personalized mentorship
This can level the playing field.
A student without mentors can now have an AI that grows with them.
Built With
- Hindsight (memory layer)
- Groq (LLM inference)
- React (frontend)
- Node.js (backend)
Try It Yourself
The project is open source and available on GitHub.
You can run it using Hindsight Cloud and a free API key.
Because the best career advisor is one that never forgets you.
Final Thoughts
This project taught me something simple:
AI doesn’t become useful just by becoming smarter.
It becomes useful when it starts remembering.
And once it remembers — it starts understanding.
Built for the AI Agents That Learn Using Hindsight Hackathon.
Top comments (0)