PyTorch Foundation Ambassadors bring people together to learn, share technical knowledge and contribute to PyTorch Foundation projects and the wider open-source AI ecosystem.

In this November roundup, we highlight some of our ambassadors’ activities, including local meetups, university sessions, technical articles and practical learning resources.

PyTorch Community Meetup in Kampala

On September 19, Pacifique Mugisho organized an in-person PyTorch East Africa community meetup at Makerere University in Kampala, Uganda, in collaboration with the IndabaX Makerere Student Chapter and the IEEE Computer Society Makerere Student Branch.

The event received 103 registrations and welcomed 32 attendees. Talks explored retrieval-augmented generation (RAG), external knowledge integration and advanced retrieval techniques. Pacifique also led a practical workshop covering embeddings, semantic search and how these components can be combined in PyTorch-based applications.

An interactive quiz gave participants another way to engage with the material. Following the event, the community gained 20 WhatsApp members and four LinkedIn members. The gathering also laid the groundwork for further in-person meetups.

Technical talks and hands-on learning at the PyTorch East Africa community meetup at Makerere University, Kampala.

Hands-on PyTorch on TPUs in Dhaka

On October 30, Zaynul Abedin Miah delivered a technical talk and live demonstration, “PyTorch on TPU with TorchTPU,” at a community meetup in Dhaka, Bangladesh.

The session showed students and early-career developers how to run native PyTorch on Google TPUs using free Colab notebooks. It covered device setup, tensor operations and torch.compile, with an interactive “predict, then run” exercise.

The event received 91 registrations, and all 50 approved seats were filled. Participants travelled from several cities, with some journeys exceeding 330 kilometres. Slides and a runnable notebook were shared with all registrants so they could continue experimenting afterward.

View the event. Colab: Link

Attendees and a live demonstration at the PyTorch on TPUs community session in Dhaka.

Efficient LLM Inference with PyTorch Community Nepal

On November 1, Arun Bhandari organized and hosted PyTorch Community Nepal’s first online technical session, “Accelerating LLM Inference with Speculative Decoding at the Edge.”

Guest speaker Manjil Nepal, a Research Associate at the Indian Institute of Science in Bangalore, discussed speculative decoding in edge–server environments and approaches to improving draft models through local and federated learning.

Arun coordinated the speaker, registration, promotion, moderation and audience discussion. The session attracted 86 registrations and 20 attendees, with presentation slides and source code shared afterward.

The event created an opportunity for researchers and Nepal’s AI and machine learning community to exchange knowledge and connect through a technical discussion.

PyTorch Community Nepal’s online session on efficient LLM inference.

AI Research with PyTorch in Nigeria

During September, Abdulsalam Yazid Bande delivered two introductory sessions for academic audiences in Nigeria, reaching 75 attendees.

“From Research Question to PyTorch Experiment” introduced 35 graduate and postgraduate students to modern AI, neural networks, generative AI and PyTorch’s role in research.

A second session, “Getting Started with AI Research Using PyTorch,” welcomed 40 faculty members for a practical introduction to experimentation, model development and machine learning applications in academic research.

Together, the sessions helped students and faculty explore how PyTorch could support their own research questions.

Introducing AI research with PyTorch to academic audiences in Nigeria.

LLM Application Development at Vilnius Tech

In September, Eyup Cinar visited Vilnius Tech University in Lithuania to deliver a guest lecture on rapid application development with large language models.

Approximately 30 undergraduate and graduate students attended. The session introduced the PyTorch ecosystem and its role in developing, training and deploying LLM-based applications, while providing an opportunity for academic exchange between institutions.

A guest lecture on LLM application development at Vilnius Tech.

Getting Started with Open Source at Duke University

On September 4, Sahdev Zala presented to more than 25 students in Duke University’s AI Engineering master’s program in Durham, North Carolina.

The session helped students understand how to get started contributing to PyTorch and Kubernetes. Participants joined both in person and online to explore opportunities to contribute to open-source projects.

Exploring open-source contributions with Duke University students.

LLM Post-Training at DevFest Kuala Lumpur

On November 14, Vensen Mu, PyTorch Foundation Ambassador and RL-Kernel maintainer, delivered “Practical Train–Inference Consistency in LLM RL Post-Training” at Google DevFest Kuala Lumpur 2026.

The talk explored numerical discrepancies between training and inference systems in reinforcement learning for large language models. Vensen shared how the open-source RL-Kernel project works toward bitwise consistency across training and rollout workflows, including engineering considerations relevant to PyTorch and vLLM.

The session encouraged developers interested in AI infrastructure and reliable RL post-training to explore the problem and collaborate through open source.

Read the event announcement and explore RL-Kernel.

Sharing PyTorch Updates in Simplified Chinese

Zhiqing Xiao published a Simplified Chinese WeChat article introducing PyTorch 2.14.

The article received more than 300 views, helping Chinese-speaking readers learn about the release.

Read the article in Simplified Chinese.

Transfer Learning: Articles, Code and Community Talks

Sahdev Zala published an IBM Developer article based on a transfer-learning demonstration using PyTorch and Torchvision’s pretrained ResNet-18 model. He also made the accompanying source code publicly available, giving readers an example to explore and build on.

Read the transfer-learning article and explore the accompanying code.

Thank You to Our Ambassadors

Thank you to our ambassadors for sharing your knowledge, creating learning opportunities and bringing people together around PyTorch.

The post PyTorch Foundation Ambassador Community Highlights November 2026 appeared first on PyTorch.