Kalyan Sekhar

Pioneered anomaly detection in user behavior patterns within Endpoint Management (EPM) by leveraging an ensemble of advanced statistical and machine learning models, including LSTM, Time Series GANs, and Isolation Forest. Successfully reduced false positive rates by 40%, significantly bolstering security measures. Developed post-processing APIs for seamless SQL database integration and integrated findings into the UI, achieving a 63% reduction in Page Response Time through optimized queries and Redis caching. Engineered a robust document search system, incorporating keyword similarity indexes and utilizing LLM models (Llama2 and GPT-2). Developed a streamlined pipeline with Langchain for training, fine-tuning which ultimately resulted in an perplexity score of 25. Designed and Deployed Pam Modules for authentication for Windows, Linux, and Unix Systems including AIX, and Solaris systems.

Developed and implemented Object Detection (Faster R-CNN) and Segmentation (Mask R-CNN and UNet) models from scratch on the DeepEdge platform. Ensured high customizability throughout the implementation and seamlessly integrated the models for enhanced functionality. Implemented MongoDB for data injection and extraction with API integration. Specialized in extracting experiment-related data, facilitating efficient checkpoint loading, and model fine-tuning. Optimized database calls and refined APIs, leading to a remarkable 200% improvement in response time. Elevated overall system efficiency for a superior user experience.