Gifty

• 

  Demo

Our App

https://giftyai.org

Inspiration

Finding the perfect gift can be stressful. With so many options available, it’s easy to feel overwhelmed or worry that the recipient won’t truly love what you choose. Many people default to generic gifts or last-minute purchases, which can feel impersonal and uninspired.

This app was inspired by the desire to make gift-giving more meaningful and effortless. By using AI to analyze the recipient’s interests, preferences, and even past gifts, the app ensures that every present feels thoughtful and well-matched. Instead of spending hours searching, users get personalized recommendations in minutes.

Another key inspiration is the emotional impact of a well-chosen gift. A thoughtful present can strengthen relationships, show appreciation, and create lasting memories. But not everyone has the time or knowledge to pick the perfect gift—this app fills that gap.

Finally, the rise of data-driven personalization in shopping made this idea possible. With AI and smart algorithms, we can tailor recommendations more accurately than ever before. This app makes high-quality, personalized gifting accessible to anyone, ensuring every occasion is special.

What it does

GiftyAI makes gift shopping effortless and personalized! Simply chat with our AI assistant, Gifty, in a quick, friendly conversation about the recipient—their interests, hobbies, and preferences.

Then Gifty dives into deep research, scanning the internet for the best products. It reads reviews, analyzes specifications, and compares options to ensure the gifts align perfectly with the recipient’s profile.

After thorough research, Gifty presents you with three top gift choices. Each option includes a detailed description, review summary, price, and a direct link to purchase—making your decision simple and stress-free.

With GiftyAI, you’ll always find the perfect gift, backed by smart AI recommendations, so every occasion feels extra special!

How we built it

We wanted GiftyAI to feel as natural and personal as giving a gift to a close friend or family member. This inspired our soft pastel pink and white color scheme, as well as the friendly design of our mascot, Gifty, who embodies warmth and helpfulness.

To make the experience even more engaging, we integrated 11Labs for voice interaction. Instead of just filling out a form, users can have a spoken conversation with Gifty while building a gift recipient’s profile. This adds a human touch, making the process feel more intuitive and real.

For product research, we use the Spider API to scrape reviews and extract website markdown from relevant sources. This data is then processed through Google’s Gemini-2.0-Flash model, which efficiently cleans and condenses large amounts of information. We pass this refined data to OpenAI’s GPT-4o-Mini, which summarizes key insights and selects the three best gift options based on the user’s input. To enhance the experience further, we use DuckDuckGo Search to find product images and Spider API with Gemini-2.0-Flash-Lite to locate purchase links.

Finally, all this data is seamlessly presented in a clear, user-friendly format, ensuring that users can confidently choose the perfect gift with minimal effort.

Tech Stack:

• Python Fastapi Backend
• NextJS Frontend
• OpenAI and GoogleAI models for LLM work

Challenges we ran into

Our original vision was to create a broad product recommendation tool, but as we developed the idea, we realized that the most difficult and time-consuming part of gift shopping is the research. This insight led us to pivot the entire project midway through the second day—a major challenge, as it meant we had far less time than planned to execute our new approach.

One of the biggest technical hurdles was reliably finding direct sales pages for selected products. Initially, our searches returned a mix of unrelated pages, making the results inconsistent. After extensive fine-tuning, prompt engineering, and data cleaning, we optimized our LLMs to consistently extract and return the exact purchase links we needed.

Another major challenge was the user experience. Our initial approach used text input for building the recipient’s profile, but it felt unnatural and tedious. One team member suggested using 11Labs for a spoken conversation instead, making the process more engaging. Implementing this, however, introduced new difficulties, particularly with managing tool-calling states and syncing them with the conversation flow.

To solve this, we used React state refs instead of just useState, ensuring a more responsive and up-to-date state when handling tool call functions. Overcoming these challenges made GiftyAI feel more intuitive, efficient, and truly personalized.

Accomplishments that we're proud of

We’re incredibly proud of the entire project, but a few aspects stand out as our biggest achievements.

First, the conversational profile-building feature was both a technical challenge and a rewarding success. Instead of a simple prompt-response system, we created a dynamic, natural conversation where Gifty listens, speaks, and even handles interruptions smoothly. The result feels more like chatting with a friend than filling out a form, making the experience truly engaging.

We’re also especially proud of our deep research functionality. Our app doesn’t just suggest generic gift ideas—it takes a detailed profile and a selected category and returns three highly relevant products, complete with descriptions, images, purchase links, and personalized justifications. Seeing this process work seamlessly has been incredibly rewarding.

One unexpected challenge was the time required for research. We didn’t want users staring at a loading screen, so we designed an interactive research page featuring Gifty running around, "thinking" as it gathers results. This playful addition keeps users engaged and enhances the experience, and we’re really proud of how it turned out!

What we learned

Teamwork: Our team started with a mix of skill sets—some of us were new to frontend work, while others weren’t as comfortable with backend development. This taught us that effective teamwork was essential to creating a polished, well-rounded product. We learned how to delegate tasks based on each person’s strengths, while also supporting each other through unfamiliar challenges.

AI: For many of us, working with AI was a completely new experience, and it presented some tough learning curves. We dove into selecting the best models for our tasks, learning about concepts like context windows and tokens. Prompt engineering became crucial, as we had to craft our inputs carefully to ensure the LLMs understood exactly what we wanted and consistently delivered the right results.

Technical: From a technical perspective, we faced numerous challenges. Some of us were new to frontend development, while others had never worked with the frameworks we chose, like FastAPI for Python and NextJS for the frontend. There was a lot of hands-on, on-the-go learning, which, although tough at times, was incredibly valuable. We not only gained technical knowledge but also honed the skill of quickly picking up new tools and frameworks—an essential skill for any developer.

Built With

• 11labs
• fastapi
• googleai
• javascript
• nextjs
• openai
• python
• react
• spider
• typescript