Aayush Raina

Developed anomaly detection systems using active learning techniques and implemented domain adaptation strategies for tuning a universal model. Designed dynamic weight initialisation methods for policy-based networks and created drift monitoring solutions using supervised and unsupervised methods while preserving anomalies with under-sampling techniques.

Designed automated online monitoring (drift detection) systems using distribution based and auto-encoder based methods to monitor and identify defects in real-time during the manufacturing process. Led the roadmap planning and execution for scaling from 50k defects to 2M defects for filtering domain by driving various proof-of- concept (POC) initiatives to validate new technologies and approaches. Played a key role in transforming a next-generation product from MATLAB to Python, considering scaling requirements and business cost implications. Analysed and resolved throughput and memory issues across application, data engineering, and machine learning algorithm modules. Contributed to big data features in datacenter for CRUD operations, data engineering, observability, distributed caching, and notification services for different user and backend flows. Engaged with customers and business lines to gather requirements, provide technical guidance, and ensure alignment with business objectives. Facilitated cross-functional collaboration to ensure successful delivery of architectural solutions.

Designed and implemented an in-memory decompression algorithm for compressed files using zlib, optimising resource utilisation and improving performance by 40%. Enhanced various device management flows by implementing indexing, caching, and priority scheduling.