About Me

Hi There!

You have stumbled upon my blog about software engineering and computer science. On this blog, I write about topics that inspire me in my work as MLOps Engineer, or the projects that I work on in my spare time. However, this page is about me, a Swiss Software Engineer from Lucerne who currently lives and works in the federal city Bern. Read on to learn more about me and my projects.

My Career

• I work in a team that provides a MLOps platform for Data Scientists. We enable them on their journey from training models to bringing them into production.

• I maintained the infrastructure for training and running AI language models with a focus on making inference more efficient and scalable.

• I worked on a prototype for a reproducible data platform in the neuroscience domain using Docker and semantic web technologies.

My Education

• I took classes and created projects in the domains of distributed computing, data science, and information security.
• I graduated with a thesis on decentralized machine learning for which I collaborated on the development of a novel algorithm for communication efficient decentralized learning.

• I primarily focused high performance computing with an interest in the intersection of machine learning and scalable computer systems.
• I graduated with a project that enabled simulating different distributions of stale gradients to analyze their impact on training ML models with asynchronous stochastic gradient descent (SGD).

Open Source Projects

• FastDynamicBatcher: a FastAPI extension for batching inputs across requests to accelerate machine learning workloads.
• CompIntPy: A Python library with C++ binding for fast encoding and decoding of variable length integer arrays.

Publications

• A. Dhasade, A. -M. Kermarrec, R. Pires, R. Sharma, M. Vujasinovic and J. Wigger, "Get More for Less in Decentralized Learning Systems," 2023 IEEE 43rd International Conference on Distributed Computing Systems (ICDCS), Hong Kong, Hong Kong, 2023, pp. 463-474