Delta | Personal AI Career Mentor

Home Page

Inspiration

As engineering students passionate about IoT, robotics, and hackathons like SIH and Robofest, we were frustrated by the scattered career advice available to peers in India. Traditional roadmaps ignore personal contexts like tight academic schedules and regional job markets, leaving students confused. Delta was inspired by Google Gemini's multimodal strengths to create a dynamic AI mentor that evolves with each user's journey, turning "What should I learn next?" into actionable, personalized plans.

What it does

Delta transforms student career confusion into structured success using five Gemini-powered flows: deep profile analysis for SWOT insights, dynamic 12-month roadmaps with curated courses/projects/internships, context-aware encouragement tied to real progress, auto-generated professional resume bullets and LinkedIn posts, and a chat-based "Roadmap Assistant" that adjusts plans for life's curveballs like exams.

How we built it

We centered Gemini 3 API (Flash for speed, Pro for depth) as the AI engine with chain-of-thought prompting for reliable JSON outputs. Python Flask orchestrates user state and API calls, SQLAlchemy with SQLite stores progress data, and a vanilla JS/CSS frontend visualizes complex JSON into interactive cards. Multimodal inputs like resumes or goal notes feed into Gemini for tailored outputs.

Challenges we ran into

JSON hallucinations broke our nested roadmaps early on—Gemini's structured output mode fixed this but required strict schema prompts. Balancing latency for chat (Flash's blazing speed helped) versus deep analysis was tricky; we also wrestled with context retention over long user sessions, solved by chunking history into prompts.

Accomplishments that we're proud of

The "magic flow" from task completion to instant LinkedIn-ready posts feels seamless and professional. Visualizing Gemini's raw JSON into an intuitive, responsive UI was a win, making complex 12-month plans glanceable. It reliably handles real student scenarios.

What we learned

Gemini excels as a "logic engine" for decisions like internship timing, not just text generation. Speed (Flash 3) unlocks conversational UX we couldn't dream of before. Prompt engineering is as critical as code—chain-of-thought ensured explainable, stable outputs every time.

What's next for Untitled

Integrate user memory for multi-year tracking, add India-specific job APIs (Naukri, LinkedIn), and support multimodal uploads like portfolio images for visual skill analysis. Open-source expansions for niche fields like IoT/robotics roadmaps, with mobile app deployment via Hostinger. Aim for production as a free tool for 1M+ Indian students.