@PyTorch: Optimizing kernels can be time-consuming and require deep expert knowledge. KernelAgent is a multi-agent harness that c…
Summary
Kaiming Cheng of Meta will present KernelAgent, a hardware-guided multi-agent harness that optimizes GPU Triton kernels, at PyTorch Conference North America 2026; it achieved a 2.02x speedup over prior versions and 1.56x over default torch.compile across 100 KernelBench L1 tasks.
View Cached Full Text
Cached at: 10/08/26, 10:06 PM
Optimizing kernels can be time-consuming and require deep expert knowledge. KernelAgent is a multi-agent harness that can write and optimize kernels for you.
At #PyTorchCon North America 2026, @KaimingCheng (@Meta) will present “KernelAgent: Hardware-Guided GPU Kernel Optimization via Multi-Agent Orchestration.”
The session will cover a hardware-guided optimization layer that integrates GPU hardware-performance signals into a closed-loop multi-agent workflow to guide optimization for Triton kernels. Across all 100 KernelBench L1 tasks evaluated, KernelAgent achieved a 2.02x speedup over kernels generated by earlier versions and an average 1.56x speedup compared with default torch.compile.
Register for PyTorch Conference North America 2026: https://hubs.la/Q04v4SL60
Register | PyTorch Conference North America: The Deep Learning Conference
**Where what’s next in open source AI gets decided.**Don’t Miss Out!
PyTorch Conference Rates
The registration deadline is 11:59 PM Pacific Time on the respective date.
Attendee
Super Early BirdMar 31–Apr 10, 2026
Early BirdApr 11–Jul 31, 2026
StandardAug 1–Sep 4, 2026
LateSep 5–Oct 21, 2026
$ 999 USD
Academic*
Super Early BirdMar 31–Apr 10, 2026
Early BirdApr 11–Jul 31, 2026
StandardAug 1–Sep 4, 2026
LateSep 5–Oct 21, 2026
$ 249 USD
Bringing a Team? Purchase 10+ registrations at one time and receive a 25% discount. Contact[email protected]to claim your discount.
*Must provide proof of academic ID
What’s Included?
Your PyTorch Conference Registration Includes:
- All conference content – keynotes, conference sessions, panels, and more.
- Social Events & Networking Activities: Flare Party
- Community Expo entry
- Lunches, morning and afternoon coffee & snacks
- Event swag
- AI Community Bash featuring DE LA SOUL
Registration Types
**Academics:**Academic and student registrations are forcurrent full-timestudents, faculty/staff members. Those who qualify will need to upload a valid copy of their institution-issued ID showing the expiration date, current class transcript, OR proof of current employment showing they are a full-time student or faculty at the time of registration. If you have any questions, please contact[email protected].
Small Business: Our small business registration discount is available to companies that have less than $5 million in annual revenue AND less than 30 employees. Please contact us at[email protected]to request the Small Business Registration discount code.
Speakers: You should have received a registration link in your acceptance email. If you did not, please contact[email protected]for more details.
Sponsors: A registration link was shared in an email to your company’s sponsorship contact. Please reach out to your company’s sponsorship contact if you need to register as a Sponsor. For further questions, please email[email protected].
Media: If you are a member of the media interested in attending this event and have not received a complimentary access code to attend, please contact[email protected]and a member of our PR team will be in touch.
Additional Information
Quick Note: We never sell attendee lists or contact information, nor do we authorize others to do so. If you receive an email claiming to sell an attendee list for a Linux Foundation event, please forward it to[email protected].
Scholarships
Learn more aboutscholarship and travel fundingopportunities.
Invoices & Certificates of Attendance
Registration Invoices: Invoice receipts are downloadable from the confirmation email you received after registering under thePayment Receipt Informationsection. If the downloadable invoice receipt does not meet your needs or you need to have your confirmation email resent, pleasesubmit a request. Please include any additional customization you need for your invoice receipt in the request.
Certificates of Attendance: You can download your Certificate of Attendance directly from yourLFX dashboardby navigating to the “Past Events” section in the LF Events column.Please Note: We verify attendance through the registration system, and certificates will only be available for download after the event is completed.
Refund Policy
Cancellations: If you must cancel for any reason, pleasesign back into your registration, click the “Register/Modify” button, and select “Unregister.” Refunds will only be issued for cancellations received two weeks prior to the main event start date (including bulk ticket invoice requests, co-located events, and add-ons), and will appear as a credit on the card’s statement 7 – 10 business days after cancellation. Refunds can only be issued on the card the original payment was made.
Please note that all refunds are subject to a6% processing fee.
Substitutions: If you are unable to attend, you may substitute another attendee in lieu of cancellation. To substitute an attendee,sign back into your registration, click the “Register/Modify” button, and select “Transfer Registration” on your confirmation page. If you need further assistance, email[email protected].
Similar Articles
KernelArc: A Multi-Agent Framework for GPU Kernel Optimization
KernelArc is a multi-agent framework that uses strategy-specialized agents to autonomously optimize GPU kernels across heterogeneous workloads, achieving top rankings on NVIDIA GPU benchmarks.
AgentKernelArena: Generalization-Aware Benchmarking of GPU Kernel Optimization Agents
AgentKernelArena is an open-source benchmark for evaluating AI coding agents on GPU kernel optimization, assessing full agent workflows and generalization to unseen configurations across 196 tasks.
@HongyiJin258: AI agents are becoming GPU kernel engineers. We're rethinking the compiler stack around them. Meet TIRx Harness: an ope…
MLC AI releases TIRx Harness, an open compiler harness that gives AI agents a structured environment to explore, debug, and optimize GPU kernels. Agent-built Kimi Delta Attention kernels achieved 2.94x geometric-mean speedups over FlashKDA (forward) and 6.84x over FLA (backward).
Kernel Forge: An Agent Harness for LLM-based Generation and Optimization of CUDA Kernels
Kernel Forge is an open-source agent harness that uses LLMs and Monte Carlo Tree Search to automatically generate and optimize CUDA kernels for any unmodified PyTorch model, achieving up to 2.83× speedup on softmax in Gemma 4 E2B.
Optimizing CUDA like a Human: Micro-Profiling Tools as Expert Surrogates for LLM-Based GPU Kernel Optimization
KernelPro is a closed-loop multi-agent system that uses LLMs and micro-profiling tools to automatically optimize GPU kernel code, achieving geomean speedups of 2.42×/4.69×/5.30× on KernelBench and demonstrating a measured 11.6% energy reduction at matched speed.