Varun Verma

Uptycs

06.2023 - Current

Engineered Ask-Uptycs, a hybrid search system combining semantic and keyword-based search, improving accuracy and relevance. Extended search capabilities to dashboards and help docs, enhancing accessibility across platforms.
Optimized search and data storage by transitioning from flat-file storage to a vector database, enabling efficient handling of large datasets.
Built an SQL assistant powered by Generative AI (GenAI) to enable dynamic querying across 250+ tables, eliminating reliance only on predefined dashboards. Added chat history storage for improved usability and continuity.
Worked on Software Cataloging, leveraging GenAI to extract software details from command-line inputs. This expanded software detection beyond a predefined dataset, improving identification in user environments.
Implemented Detection Summarization, utilizing GenAI-powered summarization to condense complex detection graphs into readable summaries, enabling quick insights into system anomalies.
Contributed to the Container Anomaly Detection project, utilizing telemetry data to identify suspicious activities in containerized environments.
Improved data management by integrating the Iceberg table format, enhancing performance and scalability.
Enhanced dashboard performance by migrating from Trino to DuckDB as the OLAP query engine, significantly reducing query latency using Alluxio-based filesystem caching.

Summary