- Senior at Rutgers University, studying computer science
- Passionate about transforming raw data into actionable insights 🌟
- Currently working on ML/AI at Tesla
- Fun fact: I enjoy experimenting with baking recipes. Banana bread is my favorite to bake! 🍞
What We Did: Developed topic modeling modules to extract key themes from online customer reviews. Created improved alternatives to an existing solution that produced excessive topics (100-300) for each product subcategory.
My Accomplishments: Implemented and refined two models: Latent Semantic Analysis (LSA) and BERTopic (with k-means clustering)
- Achieved comparable topic quality to the original BERTopic implementation using LSA, a lightweight, efficient topic modeling algorithm
- Developed topic evaluation method using sentence transformers and similarity score
Tools: Python, Gensim (LDA2Vec), scikit-learn (LSA), Top2Vec, BERTopic, Groq (to use Llama3), HuggingFace sentence-transformers
Result: Models generate 10-20 representative topics per product subcategory, enabling marketing and R&D teams to focus on critical customer feedback. Reduced topic modeling costs by using Llama3 instead of OpenAI.
Check out the full project here
- Programming Languages: Python, Java, JavaScript
- Data Science: Pandas, NumPy scikit-learn, TensorFlow, matplotlib
- AI: HuggingFace, LangChain, Ollama
- Web Dev: TypeScript, HTML/CSS, React.js, Node.js, Flask
- Databases: MySQL, Clickhouse
A few other projects I’ve worked on:
- Developed a random forest model to predict Airbnb prices in NYC with
$R^2$ = 0.6 - Tech Stack: Python, Pandas, scikit-learn, TensorFlow, NumPy, matplotlib, Seaborn
- Learnings: data preprocessing, feature engineering, model selection and evaluation, hyperparameter tuning
- A web app for real-time listing and bidding of toys
- Tech Stack: Java, MySQL, HTML/CSS, JavaScript
- Learnings: full-stack development, database integration, schema design
- Email: [email protected]
- LinkedIn: https://www.linkedin.com/in/jessica-luo336/