Founder

TasteBud

Personalized Recommendations from Noisy Dining Data

Built a conversational food recommendation experience that turns messy menu and place data into trustworthy, context-aware suggestions users can depend on.

Highlights

Hyper-personalized

Recommendations tuned to mood, taste preferences, and budget instead of generic ranking lists.

Constraint-aware

Supports dietary and allergy needs including gluten-free, diabetic, and other food constraints.

Context-sensitive

Handles situational requests like date-night patios, kid-friendly dinners, and dog-friendly spots.

Background

Context

TasteBud reduces decision fatigue by turning broad food and drink discovery into highly specific recommendations tailored to each person in the moment.

Stakeholders

  • Foodies looking for better local recommendations
  • Travelers in new places needing quick confidence
  • People with dietary or allergy constraints
  • Groups with situational needs (family, pets, events)

Challenges

  • Traditional review and map products return broad lists that often miss personal context.
  • Users need recommendations that account for diet, allergies, budget, and location at once.
  • Decision support should work across both discovery and menu-level choices.
  • Recommendation quality depends on trust, freshness, and perceived reliability of data.

Objectives

Personalized Discovery

Deliver recommendations matched to taste, mood, and budget instead of generic popularity.

Constraint Coverage

Support dietary and allergy-aware recommendations users can actually act on.

Conversational Utility

Make the product useful for natural-language questions from broad to highly specific.

Research and Discovery

Use-case Taxonomy

Product scope was structured around three primary query categories: quick recommendations, specific constraints, and menu intelligence.

Preference Dimensions

  • Mood and craving intent
  • Dietary and allergy constraints
  • Location and situational context
  • Reliability and review quality expectations

Sample Prompt Modeling

Prompts were designed to cover common scenarios like date nights, work-friendly coffee spots, vegan options, and dish recommendations at specific chains.

Design Approach

Persona-led Voice

Framed the product as a trusted foodie friend to make interaction feel human and low-friction.

Prompt-first UX

Centered interactions around natural-language requests instead of rigid filter-only experiences.

Progressive Specificity

Supported both broad "where should I go?" flows and precise dish-level decision support.

Trust Signals

Emphasized reliable, highly rated recommendations and real-time data quality cues.

Solution

Recommendation Layer

Delivers highly rated restaurant and bar recommendations aligned to personal preferences and budget.

Constraint Intelligence

Handles allergies, dietary rules, and lifestyle filters without forcing users through complex setup.

Scenario Fit

Supports context-heavy requests such as group lunches, date nights, pet-friendly outings, and family dining.

Quick Recommendations

Examples: nearby sushi, work-friendly coffee shops, romantic dinner spots.

Specific Needs

Examples: vegan patios, kid-friendly gluten-free options, late-night delivery.

Menu Intelligence

Examples: what to order at a chain, spicy options, popular menu items.

Reflections

TasteBud validates a direction I use across my work: translate complex data and intent signals into approachable product workflows people can trust in high-frequency decisions.