L.A.R.P.

• 

  Home

Inspiration We noticed a major gap at UCSD: the "hidden curriculum." While campus is crawling with resources like TESC, ACM, and the Career Center, knowing how and when to use them is usually gatekept by social circles. We wanted to build a mentor in a pocket—specifically for first-gen and transfer students—that demystifies the path to an internship without making them trade their personal data for a roadmap. What it does L.A.R.P. (Learn, Apply, Reflect, Progress) is a stateless career compass. You pick a track (like Data Science or SWE), check off your current "signals"—from coursework to part-time retail jobs—and the app generates a personalized, one-week action plan. It features a Skill Radar to visualize your gaps, a Spanish-language toggle for accessibility, and an ATS-friendly resume builder to help you immediately apply what you've learned. How we built it We used the Groq API (Llama 3.3-70b) for lightning-fast inference to map student experiences to industry skills. The frontend was built using Cursor for rapid prototyping, and the entire architecture is strictly stateless—all data lives in the browser's $RAM$ and is purged the second you close the tab. We also integrated a custom PII redaction layer to scrub sensitive info like emails or phone numbers before they ever hit the LLM. Challenges we ran into Building a personalized tool that remembers nothing was a massive hurdle. Managing complex application state without a backend database meant we had to be very intentional with how we handled user inputs. We also spent a significant amount of time fine-tuning our redactPii logic to ensure that a student’s "free-text" descriptions didn't accidentally leak their identity to the model while still providing high-quality advice. Accomplishments that we're proud of     •    True Data Sovereignty: We successfully built an app that provides deep personalization with zero persistent storage.     •    Bilingual Support: Implementing a functional Spanish toggle that actually adapts the career advice, not just the buttons.     •    The Scrubber: Building a robust PII redaction system that gives us (and the user) peace of mind. What we learned We learned that privacy is a feature, not a hurdle. You don't need a massive database of user behavior to be helpful; you just need a smart way to map their current context to existing resources. We also realized how much "hidden value" exists in non-traditional work—like how customer service experience is a massive "Unlock" for Outreach and Management roles. What's next for L.A.R.P. We want to add a "Local Export" feature so users can save their roadmaps as PDFs or Markdown files without us ever touching the data. We’re also looking into a Project Suggestion Engine that recommends specific technical projects to fill the gaps identified on the Skill Radar. Finally, we want to expand our language support to include Mandarin and Tagalog to better serve the full UCSD community.

Built With

• css
• groq
• html
• javascript
• typescript