Future Helper

Inspiration

The inspiration for this project came directly from observing the environment around me. Today's high school and university students experience immense anxiety regarding their career choices due to a rapidly changing job market and economic uncertainties. Seeing that over 40% of youths are unsure about their career goals—a statistic that directly impacts their future happiness and success—motivated me to take action. I wanted to build a personalized, data-driven compass that could guide young people through this overwhelming process, analyzing their unique strengths rather than relying on generic advice.

How we built it

FutureHelper is designed as a comprehensive decision-support system. I built the application utilizing a modern full-stack architecture, featuring React for a dynamic, user-friendly frontend, and Node.js coupled with Firebase for robust backend operations and database management. At the core of the platform is the integration of psychometric evaluations—specifically the DISC, RIASEC, and OCEAN personality tests. To process this data, I integrated the OpenAI API, which acts as the analytical engine. The system takes the student's test results, academic achievements, and life values, and computes a compatibility profile. To quantify the alignment between a student's psychometric profile and a potential career path, the system relies on vector similarity. For instance, the foundational logic for calculating a career match can be represented using a weighted cosine similarity formula: Match Score= ∑ i=1 n ​ w i 2 ​

​ ⋅ ∑ i=1 n ​ S i 2 ​

​

∑ i=1 n ​ (w i ​ ⋅S i ​ ) ​

Where S i ​ represents the student's score in a specific personality dimension (e.g., Openness to Experience or Investigative traits), and w i ​ Represents the weighted importance of that trait for a specific target profession. The AI then synthesizes this mathematical alignment into a readable, personalized career and university report.

Challenges we ran into

Developing a full-stack application from the ground up presented significant technical hurdles, particularly in managing state across React components and ensuring real-time database synchronization with Firebase. However, the most complex challenge was orchestrating the OpenAI API to generate consistent, highly structured, and hallucination-free reports based on raw psychological data. It required rigorous prompt engineering and testing to ensure the AI's advice remained objective and strictly tethered to the psychometric inputs. Additionally, preparing the project for rigorous academic scrutiny at BUCA IMSEF—and successfully qualifying for the FJSL (Luxembourg International Science Fair)—demanded that I refine not just the code, but how I communicated the social and technical value of the system to expert juries.

What we learned

Through FutureHelper, I profoundly expanded my technical toolkit, mastering full-stack deployment, API integration, and system architecture. Beyond the code, delving into the frameworks of RIASEC and OCEAN taught me a great deal about behavioral psychology and the structural gaps in current educational systems. Most importantly, presenting this project internationally has shown me the universal need for better educational technologies, reinforcing my belief in building tools that create measurable, positive impacts on people's lives.

Built With

• allowing-seamless-state-synchronization-across-the-application.-artificial-intelligence-&-machine-learning:-moving-away-from-third-party-apis
• and
• and-entirely-self-contained.-here-is-the-complete-technology-stack-that-powers-the-platform:-frontend:-react-was-utilized-to-build-a-dynamic
• and-intuitive-user-interface
• ensuring-a-smooth-experience-for-students-taking-the-assessments.-backend:-node.js-handles-the-core-server-side-operations-and-business-logic
• futurehelper?s-architecture-was-designed-to-be-robust
• ocean
• providing-a-fast-and-asynchronous-environment-to-process-user-requests.-database-&-cloud-services:-firebase-is-used-for-secure-user-authentication-and-real-time-database-management
• responsive
• riasec
• scalable
• the-core-analytical-engine-is-powered-by-a-custom-trained-machine-learning-model-developed-with-pytorch.-this-model-autonomously-processes-the-complex-psychometric-data-derived-from-the-disc