ICAN Patents
PRF Technology Reference Numbers: 2019-BAGC-68646, 2020-CHAT-68827, 2021-CHAT-69504, 2021-CHAT-69369, see additional ones below
The top two are on optimizing the end-to-end latency of serverless cloud workflows to be able to use serverless cloud technology for a complex machine learning workflows, computer vision problems, and interactive, low-latency requirements of IoT and user-centric applications. ORION focuses on performance modeling where we model latency as a distribution, rather than as a point, and on reducing cold-start latencies, especially relevant to more esoteric IoT workflows. WISEFUSE focuses on automated DAG (directed acyclic graphs) workflow transformations to horizontally bundle (bundling) parallel invocations of functions in a DAG or vertically fusing (fusion) of in-series functions of a DAG for optimization of DAG structure. It then enables the correct selection of VM sizes (on the user side) or the correct assignment of.
ORION, Optimized Execution Plans for Serverless Cloud DAG Workflows
WISEFUSE, Optimized Execution Plans for Serverless Cloud DAG Workflows
LiteReconfig: Cost and Content Aware Reconfiguration of Video Object Detection Systems for Mobile GPUs
Persistent Airborne Surveillance using Semi-Autonomous Drone Swarms
Argus: Drone-based Surveillance with Context-aware Deployment
SONIC: Application-aware Data Passing for Chained Serverless Applications
ApproxDet: Content and Contention-Aware Approximate Object Detection for Mobiles
OptimusCloud: Heterogeneous Configuration Optimization for Distributed Databases in the Cloud
Online Reconfiguration of Clustered NoSQL Databases for Time-Varying Workloads