About the Project

Inspiration

The rapid rise of TikTok as a social media and shopping platform inspired us to create a tool that leverages AI to enhance the shopping experience. I wanted to provide users with personalized product recommendations based on their preferences and trends, making shopping on TikTok more intuitive and enjoyable.

What I Learnt

Building this project taught us a lot about the complexities of recommendation systems and the importance of understanding user preferences. I delved into natural language processing (NLP) to extract meaningful insights from user inputs and learned how to integrate various machine learning models to provide accurate recommendations.

How I Built It

Data Collection: I started by gathering a diverse dataset that included various user preferences, trends, and product information.
Model Training: I used a RandomForest model to predict product recommendations based on user input. The model was trained using historical data to learn patterns and preferences.
NLP Integration: I integrated the OpenAI GPT-3.5 API to process natural language inputs, allowing users to describe their preferences in a more conversational manner.
Backend: The backend was built using Flask, which handles user requests, processes data, and interacts with the machine learning model.
Frontend: The frontend was developed with HTML, CSS, and JavaScript to provide a user-friendly interface for inputting preferences and displaying recommendations.
Deployment: The application was containerized using Docker for easy deployment and scalability.

Challenges I Faced

Solo Effort: I wasn't planning to join this hackathon due to my busy schedule with internship and school modules. However, the scale of the event intrigued me, and I decided to take on the challenge. With limited time and no team, I anticipated and faced numerous challenges, from brainstorming ideas to technical debugging. Despite these hurdles, participating in this hackathon was a rewarding experience that allowed me to apply my knowledge and learn even more.
Model Accuracy: Fine-tuning the machine learning model to improve accuracy and handle edge cases required a lot of experimentation and validation.
NLP Processing: Integrating NLP to accurately interpret user input and extract relevant information was complex. I had to iterate on our prompts and processing logic to get reliable results.
Deployment: Containerizing the application and managing environment variables securely in Docker was a learning curve. Ensuring that secrets were handled appropriately during deployment was crucial.
Data Quality: Ensuring the dataset was comprehensive and representative of different user preferences and trends was challenging. I had to clean and preprocess the data extensively.

Overall, this project was a valuable learning experience that combined multiple disciplines, from machine learning and NLP to full-stack development and deployment. I am excited about the potential impact of ShopTokRecommender in enhancing the TikTok shopping experience for users.