About

I am Meng Chen, a full stack data scientist / AI engineer who have been working on a variety of data science project in both academic and enterprise scales.

Agentic RAG ChatBot

in development

Recent development of the Large Language Models advanced the GenAI implementation in the industrial space and present a great opportunity to business to boost the productivity of their workforce and to reduce the service hurdle of their clients. I have developed a QA- and Chat-based apps in both enterprise scale and proof of concept to help business navigate in this newly emergining space. One of the projects applied Agentic RAG framework to enable users to query what they need either from internet or from specific database or just based on the exsiting knowledege of the LLM in use.

Data Science Projects

intelliRefinder

Mortgage market is trillion-dollar market, and the loan origination is the primary revenue generator for the mortgage lenders. Particularly, mortgage refinance business is one of healthy measures to constantly generate profits for those companies. Recently, non-bank (small, regional) mortgage lenders play very important role in the US primary mortgage market as they account for more than 60% of all mortgage originations. However, because most of them are small and regional, non-bank mortgage lenders have to prioritize their marketing goals and constantly optimize resource allocations given their limited financial and technological resources in order to keep themselves competitive.

intelliRefinder (demo) is a product using machine learning algorithm to categorize the mortgage refinance business opportunities in given areas. It helps non-bank mortgage lenders to actively identify refinance business opportunities in target area and would save mortgage lenders time and money for optimizing their marketing strategies. Please click here for more details.

Better or Worse?

School redistricting in Linganore-Oakdale-Urbana (LOU) area has stirred outcry from local communities in the social media. In order to understand the concerns of the local communities, Frederick County Board of Education provided an online platform for community members to express their opinions and conducted a basic statistic analyses on the feedbacks. However, their analyses lack of details about what sentiment of the local communities were and why the communities show favor/disfavor to the school redistricting plans.

In order to understand better, I created a web app BetterOrWorse (demo) using the in-house developed python module redistrict to perform data extraction, exploratory analyses, sentiment calculation, and data visualization with regard to the community feedbacks for proposed school redistricting plans. It shows that LOU communities had more positive views towards the second-round-proposed school redistricting proposals, suggesting that the latest plans had merits that satisfy local communities. By using such interactive sentiment analyses, local education administration could effectively identify where local community needs were and tailor its resources to address them, which would save time and money. In addition, this web app could potentially serve as a platform for general public recognize its neighbors’ feelings of the school redistricting studies. For details, please click here.

fusuIdentifier

Fusulinids is one of the most informative fossils present in the stratigraphic horizon. Particularly, petroleum industry relies on them for signaling the deposit layer of oil and natural gas, as specific species are indicative of productive layers. Thus, identifying fusulinids species becomes endless effort by oil industry.

fusuIdentifier (demo) is an experimental app we built to implement convolutional neural network (CNN) to automate the identifications of fusulinids species. This product intends to help petroleum industry increase the pipeline speed of oil and gas discoveries as well as academic labs improve their research productivity on the extinct organisms. Please click here for more details.