Business Problem
- OneGrand Capital works with smaller startups, primarily funding them at the pre-seed or seed level
- Since the funding ticket size is small, they work with multiple startups to help them take off their business model
- The complete process of shortlisting the applications for funding was manual, looking at the company profile, founder’s profile, and revenue targets, amongst others, before shortlisting the suitable applications to fund
- Neuralgo worked with the OGC team to identify seven parameters that 80% of the companies evaluated upon, before being shortlisted
- We then helped them build the right metrics to classify probable matches & built an algorithm to automate the complete process of shortlisting, bringing down the application shortlisting time by more than 74%
Solution
- Neuralgo worked with the team at 1GC to build the data from 3 sources
- Founder Profiles from LinkedIn, a dump of resume files submitted, other profiles’ source links
- Innovation portal Application (The portal used by 1GC for collecting applications) API(REST) to read the structured data collected during company profile creation
- Manual Research on current Technology trends
- We then performed a preprocessing on founder profiles and the Innovation portal application data leveraging Python on Amazon SageMaker
- We then performed a cluster analysis to identify keyword categories related to a particular industry, post which features engineering by using the structured data to score the company’s current and projected growth. Finally, support vector machine was used to determine the most influential parameters that contributed to the company’s success on a given time horizon.
- A Random forest model was trained using the above-processed data to shortlist future applicants, leveraging SVM, Random Forest Classification, Python, Pandas & Scikit learn on Amazon SageMaker
- The output was published as an API endpoint to be consumed by the end shortlisting application.
Architecture
Tools/Services
Outcomes
