about

my tools

Services

Cases

contact form

Contact me

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
close
a bunch of pink and white balloons floating in the air
back
2022
machine learning, fine-tuning

Corporate Model for VC News Analysis

We developed a custom Named Entity Recognition (NER) model specifically for venture capital news analysis. This model provides startups with critical insights into fundraising deals, allowing for more informed decision-making.


About

Problem

The client needed to compile a database of venture capital (VC) activity to support their product. However, raw details about deals are buried within news articles, which are unstructured, written differently by various authors, and often packed with extraneous information that’s irrelevant to a high-level overview.


Solution

  • Assembled a 15-person data labeling team to create a high-quality training dataset.
  • Trained a custom NER model capable of extracting key deal details from news articles.
  • Developed a post-processing system to verify and integrate the extracted data into the client's workflow.

Result

  • Created a unique, industry-specific dataset.
  • Achieved 95% accuracy in identifying essential deal parameters.
  • Built a post-processing tool that ensures clean, actionable data for the client’s database.
a bunch of pink and white balls on a black background

Contact form

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.