Rohit Vernekar

Master's

Data Science

AERA TECHNOLOGY

Machine Learning Engineer

• Crafted an innovative tool utilizing Kaniko for seamless deployment of machine learning models on Kubernetes, enabling real-time inference capabilities. The deployed model boasted dynamic batch inferencing capabilities to handle data of all sizes, accommodating both Python and R models. • Orchestrated the integration of prebuilt runtime Docker images, substantially slashing deployment time for these models from 10 minutes to an impressive 30 seconds. • Engineered a data validation and drift detection feature using the Alibi library. This addition provided insightful summaries of user-provided data, elevating data quality and ensuring consistency. • Worked on an advanced application leveraging Autosklearn’s power to autonomously generate optimal ML models tailored to specific datasets. These models were seamlessly deployed on Kubernetes for real-time inference, amplifying efficiency. • Conceived and executed an efficient Async worker framework from scratch, employing the asyncio library. This alternative to Celery expertly offloaded IO tasks to dedicated workers, resulting in a remarkable enhancement in application performance. The optimization of resources was further realized as this worker adeptly handled multiple concurrent tasks previously managed by various workers. • Pioneered the creation of an HTTP client wrapper proficiently accommodating both synchronous and asynchronous requests, augmenting versatility and responsiveness.