Students of Software Engineering Research Center (SERC) bagged best student poster awards at the 14th Innovations in Software Engineering Conference held in Bangalore from 22 – 24 February.
- Rudra Dhar, Ph.D student working with Dr. Karthik Vaidhyanathan and Vasudeva Varma for his work on Leveraging Generative AI for Architectural Knowledge Management. Here is the summary of the research work as explained by the authors Rudra Dhar, Karthik Vaidhyanathan and Vasudeva Varma:
While documenting Architectural Knowledge (AK) is crucial, it is frequently neglected in many projects, and existing manual tools are underutilized. To address this, automated tools for efficient AK extraction and documentation is essential. Even after generating AK, navigating through vast the Architectural Records can be overwhelming. Hence, we propose an automated Architectural Knowledge Management (AKM) approach using Information Extraction and Generative AI, which generates AK from various source for a given system and answers architectural queries with respect to the given system.
- Shubham Kulkarni, MS by research student working with Dr. Karthik Vaidhyanathan for his work on AdaMLS: A Self-adaptive Approach for Enhancing the Quality of Service of ML-enabled Systems. Here is the summary of the research work as explained by the authors Shubham Kulkarni, Arya Marda and Karthik Vaidhyanathan:
Recent advancements in machine learning have propelled the rise of Machine Learning-Enabled Systems (MLS). Despite their potential, ensuring consistent Quality of Service (QoS) amid various run-time uncertainties remains a challenge, with uncertainties stemming from ML model variability, software components, and environmental factors\cite{b3}. To address this, we introduce the Machine Learning Model Balancer, a mechanism for dynamic model switching that prioritizes QoS. Depending on real-time needs, it can favor speed or accuracy. Building on this, we present AdaMLS\cite{b2}, a novel self-adaptive approach that extends the traditional MAPE-K loop for continuous adaptation. Through lightweight unsupervised learning, AdaMLS optimizes model switching, with a focus on maximizing a utility function considering model confidence and response time. Our object detection exemplar highlights AdaMLS’s ability to consistently achieve optimal QoS.
March 2024