Yaswanth Seela

Developed an Auto-ML tool within a Low Code/No Code SaaS platform using the Vaex library. Designed, developed, and tested API endpoints for the Mobius Low code/No code platform using FastAPI. Developed a SQL query generator using SQLAlchemy, Langchain, and OpenAI GPT-4. Leveraged Hugging Face's open-source large language models Llama 7B and nSQL-6B to automate the generation of SQL queries.

Built a predictive model that estimates a customer's propensity to churn in a future period based on the historical characteristics and behaviour. The model also helped in identifying factors that drive 'churn' (cancellation of subscription). Used advanced classification algorithms such as Random Forest, Adaboost and Logistic Regression to classify customer subscriptions. Used SMOTE sampling technique as the data set had 'churns' as a very small minority class. The churn prediction model helped the client to identify top-100 'at risk' customers with 76% precision.

Worked on the POC part of Sisense - a Data Visualization tool. Build various dashboards in Power BI.

Been a part of the fraud detection team that was responsible for the detection of fraudulent activity. Performed various tests where there's a chance of fraud and detected all the possible fraudulent activities and presented the insights in Power BI resulting in a 13% save in expenditure for the client. Built a predictive model that identifies fraud that helped client to restructures certain policies in their organization.