MyNextDate

Home Page <3
Recommendations Page <3
User Analytics <3
User Location and Similar Users Analytics <3
Explore Different City Analytics <3
Login Page <3

Inspiration

It’s 11:47 AM.

You get the text: “So, where are we going for our date this afternoon?”

You check your calendar. Nothing. Today had completely crossed your mind.

Suddenly, the pressure hits. It’s not just about picking something, it’s about picking the right thing. Too boring and it feels low-effort. Too expensive and it feels forced. Too intense and it’s awkward. Too safe and it’s forgettable. Outdoor? Indoor? Casual or romantic? What should be a thousand possibilities suddenly feel like none at all.

That moment of panic, the last-minute scramble when it has to be perfect, was the real catalyst for MyNextDate.

Modern dating offers endless options but zero clarity. We built MyNextDate to eliminate decision anxiety and replace it with intelligent, tailored recommendations that match the vibe, budget, energy, and dynamic of the relationship. We’re not just suggesting something we think might work — we provide choices designed to be perfect.

MyNextDate turns that stressful “uhhh…” moment into a confident decision.

What it does

MyNextDate is an AI-powered date recommendation engine that transforms uncertainty into optimized, personalized plans.

At its core is a vector database that stores every date activity as a high-dimensional embedding. Each activity, whether it’s a rooftop dinner, late-night bowling, or sunset hike, is converted into a numerical representation that captures its tone, energy, and experiential meaning. This allows the system to search by vibe, not just keywords.

When a user inputs preferences such as budget, indoor versus outdoor, romance level, time commitment, and energy, those preferences are embedded into the same vector space. The system performs a similarity search to find the closest matching activities.

Each activity is also scored across structured numerical dimensions, allowing deterministic filtering and weighted ranking. The final recommendation blends semantic similarity, structured constraint matching, and city-specific performance data to deliver reliable results.

Users can submit their own date ideas, which are automatically embedded and scored, allowing the platform to continuously grow smarter and more localized. MyNextDate also incorporates city-specific insights, surfacing the types of dates that historically perform best in different locations across the country. The result is not just a recommendation engine, it’s a dynamic, data-informed matchmaking layer between people and experiences. Users no longer have to guess — they get suggestions that align with preference, context, and proven success.

How we built it

We built MyNextDate as a hybrid AI recommendation system that combines structured modeling, vector embeddings, and real-time analytics.

Frontend: React + Tailwind CSS provides a clean, responsive dashboard where users can view, add, and rate past dates.

Authentication & Storage: Supabase manages user accounts and date history, with row-level security ensuring data privacy.

Vector Database: Actian Vector AI stores 750 pre-scored date activities as 9-dimensional vectors (cost, indoor/outdoor, effort, social density, time of day, duration, planning, energy, creativity). User preferences are embedded into the same space, and cosine similarity searches find the closest matches. Repeat weighting prevents the same activity from showing too often.

Recommendation Engine: A preference vector is computed from successful past dates, weighted by rating, and queried against the vector database. Top matches are returned instantly. Users can also add custom activities, which are automatically embedded and scored.

Analytics: The dashboard shows success rates, average ratings, and trends such as “You tend to prefer cheaper, outdoor, afternoon dates,” updating dynamically as users rate new dates.

Optional Features: Free-text matching for custom dates and a “breakup button” that inverses preferences to show the worst date options.

Tech Stack: React + Tailwind (frontend), Python + FastAPI (backend), Supabase (auth & storage), Actian Vector AI DB (recommendation engine).

This setup let us deliver a fully functional, data-driven date recommendation system in 36 hours with personalized insights and instant suggestions.

Challenges we ran into

The biggest challenge was designing and building the database and algorithms to find the best matches. We needed to represent subjective experiences like vibe, effort, and romance as numerical vectors, then combine them with structured attributes to produce reliable recommendations. Balancing semantic similarity, structured constraints, and repeat weighting took careful iteration to make sure the system felt intuitive and consistent.

Accomplishments that we're proud of

In just 36 hours, we built MyNextDate, a fully functional AI-powered date recommendation app unlike anything currently available. No other platform combines structured modeling, semantic embeddings, and user analytics to personalize date suggestions with this level of precision. Beyond just recommendations, the system collects specialized, structured data on user preferences, behaviors, and city-specific outcomes, which is incredibly valuable for understanding what truly works in dating scenarios. We created a responsive dashboard where users can track, add, and rate dates while integrating city-level insights and analytics to deliver actionable, tailored suggestions.

What we learned

This project taught us how to translate human preferences into numerical vectors and leverage them for precise, personalized recommendations. We learned that combining semantic similarity with structured scoring creates far more accurate and intuitive suggestions than traditional filtering methods. Collecting specialized data not only improves the algorithm but also provides insights into broader patterns in dating behavior, revealing trends that were previously unquantifiable. Rapid iteration under tight time constraints helped us refine both algorithms and user experience, demonstrating the real-world value of data-driven decision-making.

What's next for MyNextDate

Looking ahead, we plan to expand MyNextDate into a mobile app, giving users real-time, on-the-go recommendations. We aim to broaden the platform beyond romantic dates to include family, friends, solo activities, and school trips. Future versions will incorporate user feedback loops to refine recommendations when a date doesn’t go as planned, integrate calendar syncing and reservation APIs, and provide deeper analytics to reveal trends and optimize future suggestions. The specialized data we collect will continue to grow in value, enabling even smarter recommendations and insights for users, researchers, or anyone interested in understanding human preferences and experiences.