TrendCart

TrendCart
Using the power of LLMs and Embedding to power your product search
Ultra personalized recommendations that go beyond the surface level
Product recommendations that goes deep
The underlying data of our recommendation engine (and how the foundation is laid)

Team Members: Shirin, Nabil Ridhwan, Bryan, and Pavanaa

Disclaimer: Since the app in the TikTok Dev Dashboard hasn't been marked for review, the login functionality will not work for normal people. For demo purposes, authorized judges or TikTok's TechJam Committee can contact our team leader or refer to the submitted write-up for more information.

Inspiration: TrendCart emerged from our fascination with the evolving landscape of online shopping. With a surge in digital commerce, we were inspired to create a solution that enhances user experience through personalized recommendations, addressing the shortcomings of traditional methods.

Learning Experience: Building TrendCart has been a journey of discovery and innovation. We delved into the intricacies of Large Language Models (LLMs) and Vector Databases, gaining insights into how these technologies can revolutionize product recommendations in e-commerce.

Project Development: TrendCart leverages cutting-edge LLMs to decipher complex user preferences and behaviors. By recording interactions such as product views, wishlists, and purchases, our system generates precise recommendations tailored to individual tastes. This approach not only enhances user satisfaction but also drives customer retention through personalized shopping experiences.

Challenges Faced: Throughout development, we navigated several challenges. Integrating LLMs effectively required us to optimize performance while ensuring scalability. Addressing privacy concerns and the interpretability of our model outputs were crucial considerations, demanding robust solutions to maintain trust and transparency.

Technical Implementation: The backbone of TrendCart lies in our efficient Recommendation API, supported by PineCone's Vector Database for high-dimensional data handling. This architecture enables real-time, seamless integration across platforms, ensuring rapid retrieval and dynamic updates for enhanced user interactions.

Future Enhancements: Looking ahead, we aim to enhance TrendCart with advanced features like RAG or fine-tuned LLMs for transparent model outputs. Implementing a human-in-the-loop approach will empower users to refine recommendations, while a questionnaire-based user experience will deepen insights into personalized shopping preferences.

Conclusion: TrendCart represents our commitment to innovating within e-commerce, providing a scalable, personalized shopping solution that adapts to user needs. Explore the future of online shopping with TrendCart—where technology meets tailored experiences.

GitHub Organization: Spring Onion Lovers
Frontend Repository: TrendCart Frontend
Backend Repository: TrendCart Backend
Recommender API Repository: Recommender LLM