2024-08-27

Agenda

  •  robotics use case write up slides @Jeff Brower  @Haruhisa Fukano  10 min

  •  AI4D Meeting count down preparation @Haruhisa Fukano  @vijay chintha 10 min

  •  Work stream 3 SPEAR @Borui Li (李博睿)  @Wilson Wang 10 min

  •  Work stream 5 Edge Gateway @Tom Qin  10 min

  •  AI Agent heads up @Wilson Wang   5 min



Attendees

  1. @Tom Qin 

  2. @Wilson Wang 

  3. @Borui Li (李博睿) 

  4. @Moshe Shadmon 

  5. @Jeff Brower 

  6. @John Freier  jefreier@berkeley.edu 

  7. @Qi Tang 

  8. @Victor Lu 

  9. @saiyan68 

  10. @Yona Cao 

  11. Gary Qi

  12. @Haruhisa Fukano 

  13. @Tina Tsou 

  14. @hao xie 

  15. @vijay chintha 

  16. Qingkun Li



Recording

https://bytedance.us.larkoffice.com/minutes/obusd8wu4xnc675ef4ue9k5e



Summary


The meeting discussed various aspects related to the progress and planning of different projects and work streams, the main contents included:

  • Robotics Use Case: Jeff and Fucanos are working on a general overview of the robotics use case and will reference it in relevant work streams. Tina suggests making it more specific and sharing it with interested hackers.

  • AFOD Meeting: HaruhisaFukano reported on the status of CFP submissions for the AFOD meeting, with 13 session presentation submissions and 3 Lightning talk submissions so far. There are enough time slots for the current submissions.

  • Hackerson: Yona and Gary Qi discussed the hackerson event, including how to use wegrade and how to judge the competition. They will work on criteria and a tutorial for wegrade and present it at the event.

  • Work Stream 3: Wilson Wang announced the new name for the project, Spear, and discussed its architecture with Borry Li. They plan to start coding and set up accounts for contributors.

  • Work Stream 5: Tom Qin introduced the work stream 5, edge gateway, which aims to streamline the process of discovering devices at the edge and support more protocols, especially ROS2. They will leverage the existing project Akri for device discovery.

Chapters


00:21 Discussions on Robotics Use Case and Meeting Recordings in the Meeting

This section is a meeting transcript mainly about the robotics use case. Tina Tsou initiated the discussion and asked for sharing the agenda. They talked about its overview, possible applications, and the need to upload related slides. They also discussed the availability and accessibility of meeting recordings and moved on to the AFOD meeting countdown.


11:36 Discussion on Session Presentations and Lightning Talk Submissions and Slots

This section is about the status of CFPs for a session. As of August 27th, there are 13 CFPs for session presentations of 30-40 minutes including QA, and 3 for Lightning talks of 10-15 minutes including QA. They have 15 time slots and 16 submissions. It's not clear if all can fit as submissions may still come in today.


15:57 Discussion on Hackson Preparation and Wave Grade Usage

This section is about various discussions. Tina asks Jeff to update the wiki page. They talk about the hacker song and screen sharing. Gary introduces added content on hacks and discusses applying features like wave grade. Yuna will be online and will give a 20-minute presentation on wave grade. There's also talk about awards and whether to combine or separate them for different products.


23:04 Discussions on Criteria and Titles for Hacker Song Event

This section discusses the hacker song event. Issues include awarding the first three teams, providing criteria for judging, determining a suitable title, and considering special prices. Five senior industry judges are involved. Gary is tasked to draft criteria with Yona's review. The title might be changed to be more general and highlight certain aspects.


30:16 Updates and Discussions on Projects and Architectures

This section covers various discussions. Gary Qi talks about helping with AI agent building and giving product credit on completion. Cao Yu offers online support for usage questions. Tina Tsou assigns review tasks and mentions project names. Wilson Wang discusses finalizing the name 'spear' for a project, architecture, and code contribution setup. Li Borui presents an architecture with user space and kernel space components borrowed from the operating system for an Edge Agent Runtime.


37:29 The Architecture and Components of Sphere Project for Edge Agents

This section focuses on the architecture of a system with user and kernel spaces. Key points include system agents generation, agent lifecycle, and components like spear drivers with agent runtime and toolboxes. The system aims for edge agent runtime, considering geographically distributed edge devices. It's a preliminary project architecture with ongoing work for optimization. There are discussions about toolboxes and deployment on devices.


44:48 Discussion on Project Shifu's Edge Gateway and Terminology Clarification

This section mainly discusses terminologies, project progress, and work stream 5. The terms 'device' and 'server' aren't differentiated. A detailed description of terms and architecture will be posted. Progress includes potential student code contributions. Work stream 5 focuses on edge gateway of Project Shifu, with objectives like device discovery, protocol support, and leveraging Akri for workload deployment.


52:21 Updates and Discussions on AI Agents and Edge Gateway in the Enterprise

This section discusses the update on the AI agent by Wilson Wang. It mentions real use cases in the enterprise environment and plans to implement different AI agents. Tina Tsou looks forward to it. There's also talk about Edge Gateway, suggesting putting the YouTube link in the wiki. Jiayi is on a pre-honeymoon, and How is identified as a person to work with while waiting for her return.


01:02:25 Discussions on Smart City Data Modeling and Updates

This section discusses various aspects of a project related to data systems and machine learning models. It covers data for water, sewage, and electrical systems, the use of forecasting and anomaly detection in water systems, the need to clarify data schema with smart city personnel, federated learning on edge and main nodes, involving John in model creation, and updates on work streams and Slack channels.