Tensor Talks: Initialized

Talks

Metro Manila

Saturday, tentative

About Tensor Talks

Tensor Talks is PyTorch Philippines' technical talk series on how modern AI systems are built: training, inference, the mathematics underneath, and agentic systems. It gives Filipino students and practitioners a recurring, technically deep venue, taught by local speakers.

Initialized is the first edition: an afternoon of full sessions and lightning talks, with the last hour for networking with the speakers. The series runs every two months and is run by community volunteers.

  • 1:00 to 6:00 PM Time
  • 12:30 PM Doors open
  • Free Admission
  • 60 Attendees planned

Registration

This is not first come, first served. Seating is limited, so we review every registration and select attendees based on their GitHub or Hugging Face profile. Anyone not selected at first stays on the waitlist. Up to four places are held for the venue host's team.

Call for proposals

Overview

Talks on the program come from this open call, apart from one given by the venue host. We want to hear from engineers, researchers and students who train, serve or build on models with PyTorch and its ecosystem.

First-time speakers are welcome.

Dates to remember

  • CFP opens:
  • CFP closes: at 11:59 PM, Philippine time
  • Speakers notified:
  • Confirm your talk by:
  • Schedule published and registration opens:
  • Registration closes:
  • Lightning talk slides due:
  • Session slides due:

The January dates move with the event's day if it changes.

Tracks

  • Training and Post-Training: Model training, fine-tuning, and alignment using PyTorch and related open-source stacks. Topics include distributed training, supervised fine-tuning (SFT), reinforcement learning, preference optimization, and improvements in training efficiency, stability, and resource utilization.
  • Inference and Deployment: Deploying models as reliable, continuously running services. Topics include LLM and multimodal inference, quantization, speculative decoding, KV cache optimization, cluster scheduling, and trade-offs among latency, throughput, and cost across cloud, on-premises, and local deployments.
  • Math and Foundations: The mathematics underneath modern AI systems. Topics include optimization and automatic differentiation, the linear algebra and probability that models rely on, and how architectures such as attention work.
  • Agents and Applications: Agentic systems and applications built on models. Topics include tool use, planning and memory, retrieval, evaluating agents, and case studies from real use.
  • PyTorch Ecosystem: Tools and community contributions built on PyTorch and related open-source projects: vLLM, DeepSpeed, Ray, Helion, Safetensors and others. Topics include domain-specific libraries, development and evaluation tools, and open-source collaboration.

You also pick a level for your talk: Beginner, Intermediate or Advanced.

Submission types

  • Session (20 to 30 minutes, up to 3 speakers): A focused talk covering a specific topic, concept, or case study, including Q&A.
    • Note: a talk with three or more speakers can't have only men on stage. This follows the Linux Foundation rule used at PyTorch events.
  • Lightning talk (10 to 15 minutes, up to 2 speakers): A fast-paced, high-impact presentation that delivers key insights or ideas in a concise format. There is no Q&A on stage; you take questions at the speaker corners during networking. A lightning talk is a good way to give your first technical talk.

The final length depends on how many talks we accept. Your acceptance email will state it.

Important notes

  • Submit up to two proposals. Each person can have at most one accepted talk.
  • This is a community event, so no product or vendor sales pitches.
  • Write the proposal yourself. Proposals that lean heavily on AI-generated or templated text rarely show the technical depth reviewers look for.
  • All speakers agree to follow the Linux Foundation Events Code of Conduct.
  • Strong proposals that don't fit January may be offered a slot at Tensor Talks in March 2027.
  • We take photos during the event. The submission form asks whether we may also record your talk.
  • Speakers receive a printed certificate.

How proposals are reviewed

A program committee scores every proposal from the call with speaker names hidden, on relevance to PyTorch and ML practice, technical depth, clarity and originality.

It then builds the program from the top-scored proposals, balancing tracks and levels.

More events

Meetup

ML & Coffee

Short talks and a networking session for people interested in machine learning.

Saturday, Tentative February 20, 2027

Hackathon

Torch Trials

A hackathon where teams run experiments, benchmark results, and present their findings.

Saturday, Tentative March 20, 2027

PyTorchPH on GitHub

The source of this website and our community files. Contributions are welcome.

Go to GitHub

Campus Labs

University chapters across Luzon, the Visayas, and Mindanao.

Meet the Campus Labs

Community page

Announcements, event posters, and updates from PyTorch Philippines.

Go to Facebook