ABRAHAM MAHARAJU

Developed a code generation system that intelligently generates code for user inputs. The system involved scanning repositories and storing generated prompts for the code that exceeded a predefined threshold in a prompt library. Enhanced application accuracy by 20% through seamless integration of a vector database integration that streamlined prompt retrieval within the prompt library by calculating minimal distances between user input embeddings and prompt embeddings inside prompt library. Reduced errors by 15% through the implementation of intent recognition feature which distinguishes code generation requests within user input.

Worked in building drift management application where continuous monitoring for ML models at all stages of the ML lifecycle. Utilized key performance indicators (KPIs) to calculate and detect drift in models, ensuring ongoing model performance and reliability. Conducted analysis on Concept drift and used methods involving time-series data to figure out when and how this happens. Improved model accuracy by 15% through concept drift analysis and timely model updates. Developed metrices to calculate Object detection and Image segmentation models.

Communicated results and ideas to key stakeholders at Vee. Implemented new statistical or other mathematical methodologies as needed for specific models/analyses. Optimized joint efforts between Data Engineering, business partners, UX, and People Science teams at Vee to drive business results. Able to create SQL and Python programming modules for custom insights required by our Product Teams. Leverage statistical analysis to understand what is acceptable versus outliers. Created automated KPI dashboards and monitoring reports.