I'm planning to attend MLSys 2026 this year in Bellevue, Washington.
MLSys, short for the Conference on Machine Learning and Systems, is one of the main conferences focused on the intersection of machine learning and computer systems. The 2026 conference will be held at the Hyatt Regency Bellevue from May 18 to May 22, 2026, with conference sessions running May 19 to May 21 and Industry Day on May 22.
I'm especially interested in attending because MLSys sits close to the work I care about: building reliable, scalable, and efficient AI systems that can move from research into real production environments. As AI systems become more complex, the hard problems are no longer only about model quality. They are also about training efficiency, inference cost, distributed systems, monitoring, debugging, security, reliability, and how these systems behave under real workloads.
MLSys focuses on exactly that intersection. The conference covers topics such as efficient model training and serving, LLM training and inference, compound AI systems and agent systems, ML compilers and runtimes, distributed and federated learning, privacy and security, testing and monitoring for ML applications, specialized hardware, hardware-efficient ML methods, and benchmarks and tooling for machine learning systems.
One thing I appreciate about MLSys is that it brings together both academic and industry perspectives. For people working on cloud infrastructure, reliability, developer tooling, or applied AI, that makes the conference especially relevant.
This year's conference also includes a Young Professionals Symposium on May 18, designed for undergraduate students, graduate students, and early-career researchers in industry and academia. The symposium includes keynotes, invited talks, panels, and poster sessions, which makes it a useful entry point for people who are still building their research and technical network in ML systems.
For me, I'm hoping to use MLSys 2026 to learn more about a few areas:
How teams are improving LLM training, fine-tuning, and inference efficiency
What the current research says about agent systems and compound AI systems
How ML systems are being tested, monitored, and debugged in practice
Where hardware, compilers, and runtimes are changing the cost and performance of AI workloads
How industry teams are thinking about production ML systems, reliability, and scaling constraints
I'm also looking forward to meeting researchers, engineers, builders, and operators working on these problems from different angles. Some of the most useful conference conversations happen outside the talks, especially when people compare what works in papers with what breaks in production.
MLSys feels especially timely in 2026 because the AI ecosystem is moving quickly from model demos to real systems. The next wave of progress will depend on better infrastructure, better evaluation, better observability, and better ways to make AI systems efficient and reliable. That is why I'm excited to attend and learn from the community.
If you're also attending MLSys 2026 in Bellevue, I'd love to connect.
Top comments (0)