Deloitte survey report of India Banking Fraud inspired us to do something about Indian digital consumers. As most consumers report a fraud on social media platforms and Twitter is popular among them. With this, we thought to build a public analytics platform to track fraud and issues related to Indian digital consumers.
What it does
Oraika collects text data from various data sources (Twitter, App Store, and Play Store), then analyze collected data using AI models, and then send extracted structured data to a database for dashboarding and alerting purpose.
How we built it
It is built using our open source tool Obsei which provides easy-to-use connectors to fetch text data for example Tweet V2 search API connector. For analysis purposes, it uses custom fine-tuned NLP models (sentiment, translation, classification, and embedding). Analyzed data to insert into PostgresSQL. For dashboarding Grafana and Metabase are used. The front-end is written in react framework and the back-end API system is in Python. This whole setup is deployed on AWS cloud using terraform infra scripts.
Challenges we ran into
We wanted to do historical data analysis but V2 API doesn't give us access to archived tweets.
Accomplishments that we're proud of
We are able to detect a loan related to fraud and hope we are able to increase awareness about such issues by sharing on public platforms. This in turn benefits the public in not falling prey to such modus operandi.
What we learned
To provide a measurable outcome in time bound manner.
What's next for Oraika
We like to launch a close beta to people who can help us improve the platform and then open it up to the public. Also starting public education initiatives regarding fraud patterns.
Platform access details
Please readme of GitHub repo https://github.com/obsei/oraika-chirp-web#readme