AI Auto Email Follow Up
Purpose: Following up on job applications was a tedious yet important task. To address this, we implemented an automated system that sends follow-up emails after three days if no response is received.
Work: If an email you sent to recruiters or outreach hasn’t been responded to for three days, GPT will generate a follow-up draft based on your past conversation for you.
Results: Over 60 people signed up and the demo received 40,000+ views.
Features
1. An Email label to filter recruiters vs non-recruiters
2. For labeled emails that haven’t been replied to for three days, a GPT draft a follow-email based on your prior conversation.
3. The draft sends it to your inbox to remind you to follow up, and you can customize it yourself before sending it.
Understand How English Fluency Affects Grading in Minerva
Purpose: My school, Minerva, comprises non-English native students from 40+ countries. Now, if a classmate speaks like Shakespeare, but I speak simply like a kindergartener, will the professor bias their grading?
Work: This project examines the potential bias in students' grades due to English fluency through NLP analysis. With techniques such as topic modeling, clustering, hypothesis testing, and neural network analyses, I investigated the influence of word choice, readability, and language sophistication on professors' evaluations of students' knowledge. You can view the entire project: here.
Results: Positive and significant grade-fluency correlations for all colleges, with the least significant for the Computer Science college!
Founding AI Engineer @Learn.xyz
Purpose: The content and image quality of mini courses generated by AI have been low.
Work:
- Deployed the entire backend pipeline, integrating Firebase and GCP, leveraging asynchronous processing, content moderation, and error handling.
- Engineered a real-time Slack integration incorporated with the Firebase backend, enabling the internal team to moderate and delete unethical content.
- Integrated vector embeddings on Pinecone using Retrieval-Augmented Generation techniques to enhance content quality and optimized a LLM.
Results:
- Ensure the quality of 60K AI-generated courses.
- Reduce in a 75% reduction in course creation.
Data Science @Meta
Purpose: The Metaverse requires data representation for its construction. However, numerous duplicate fuzzy data exist.
Work: Built a deduplication data pipeline using the Levenshtein algorithm and subsampling.
Results: Accelerated runtime by 20x and Mark Zuckerberg is happy. Nothing to show because it’s NDA. 😉
Data Engineering/Science @Jubo
Purpose: The designers, CEO, and customer success team lack insights into product performance and metrics for design effectiveness.
Work:
- Develop product analytics metrics for monitoring client performance.
- Created Airflow automation process for dashboards.
Results: I visualized insights with facility health metrics for 84 product modules with 1M+ data points saving 90% report time, reducing customer churn, and measure design effectiveness.
