Date Night

The interesting thing about Date Night as an engineering problem is that it is a constrained recommendation problem, not a search problem. A generic listings app answers "what restaurants are near me" and hands back everything; Date Night has to answer "what should the two of us actually do tonight" given a budget, a distance the couple is willing to travel, and tastes that differ from couple to couple. Those three constraints interact: the perfect spot outside the radius is a wrong answer, and so is the ideal-distance option at twice the budget. The recommendation engine was built to weigh all three together rather than filter on them one at a time, which is the difference between a shortlist that feels chosen and a list that feels queried.

Preferences are the input that makes or breaks a product like this, so the app treats taste as data it collects deliberately. Stated preferences feed the engine directly: what a couple tells the app about the kind of evening they enjoy shapes what surfaces, and saved favorites give the product a growing picture of what a good suggestion looks like for that particular pair. The goal is that the third plan the app proposes is better than the first, because by then it knows more about who it is planning for.

Location-awareness sounds like a checkbox and is actually a correctness requirement. A suggestion is only realistic if it is genuinely reachable tonight, so every recommendation is grounded in where the couple is and how far they said they would go. That keeps the output honest: no aspirational destinations three hours away, no filler pulled in to pad the list. The options that appear are ones the couple could stand up and go to.

Presentation carries the last mile of the product. The same engine output lands differently depending on how it is framed, so suggestions arrive organized into curated moods: romantic spots, fun activities, hidden gems. That framing does the work a concierge would do, turning raw candidates into a browsable answer to "what kind of night do we want," and the clean, focused mobile experience keeps the decision fast. The measure of success is a plan in seconds instead of an evening lost to scrolling. Saved favorites close the loop: a plan that worked once is one tap away the next time the same question comes up, and it feeds the engine's picture of that couple at the same time.

Date Night runs on Flutter for the mobile app, with the recommendation side built in Python behind a Node.js backend and the platform hosted on AWS with Firebase. Founder and CEO Joshua Doyle summed up the build this way: "There wasn't a single point where any of my requests were deemed impossible or not doable."