Tim Li

81 posts

Tim Li

Tim Li

@TimLi_DR

Co-Founder & CEO @DeepReach_ai | Building next-gen AI data & evaluation platform | Bridging global expert network with enterprise AI solutions

San Jose, CA เข้าร่วม Eylül 2025
267 กำลังติดตาม42 ผู้ติดตาม
Tim Li
Tim Li@TimLi_DR·
Physical AI isn’t bottlenecked by models. It’s bottlenecked by data representation. Great night in Palo Alto with folks from @nvidia , @DeepReach_AI , @OpenAI, @xAI, @Meta, @UCBerkeley, @amazon , @Apple and @Stanford. The consensus: To reach General Purpose Robotics, we need to move past messy physical signals and toward structured "Physical Tokens." We’re building that layer at DeepReach. Key takeaways: 1/ EgoScale is the blueprint. Scaling egocentric (POV) video is the shortcut to world-model priors. 2/ Data > Hardware. Dexterity is a representation win, not just a gripper win. (Shoutout to RealHand for the robotic piano finale). 3/ WorldBase is live. We’re open-sourcing 20k+ hours of ego-video & 40M miles of perception data to fuel the community. The "GPT moment" for the physical world is being built now. Follow me, more robotics deeptech meetup in bay area are coming!! #PhysicalAI #Robotics #NVIDIA #DeepTech
Tim Li tweet mediaTim Li tweet mediaTim Li tweet mediaTim Li tweet media
English
0
0
0
9
Tim Li
Tim Li@TimLi_DR·
US Congress just made one thing clear: AI competition is no longer just about models, it’s about data supply chains, control, and security. In Physical AI, this matters even more. Robots don’t learn from the internet. They learn from real-world, egocentric video data loops. At DeepReach.ai, we’re building: → US-based data engine for Physical AI → Trusted, scalable data pipelines → 100K+ hours/month egocentric data processing The next AI winners won’t just train better models. They’ll control the data + deployment loop. #PhysicalAI #Robotics #AI #WorldModel #Data #GTC #Nvidia #DeepTech here is the video link ---->youtu.be/Q1Woplm0HwM
YouTube video
YouTube
English
0
0
0
19
Tim Li
Tim Li@TimLi_DR·
Great embodied AI discussion today hosted by South Park Commons with @Thom_Wolf . A few big takeaways: • robotics is approaching its “ChatGPT moment” • the real bottleneck is still data • human-robot interaction is hugely underexplored • the future won’t be only humanoids • open ecosystems will accelerate robotics innovation Physical AI is getting real. #EmbodiedAI #Robotics #PhysicalAI #RobotLearning #AIInfrastructure
Tim Li tweet mediaTim Li tweet mediaTim Li tweet media
English
0
0
0
18
Tim Li
Tim Li@TimLi_DR·
American is facing structural labor supply crisis in agriculture and factores, this a great demo how robot can address that problem. but the navigation and model training require massive data in real working environment, and make sure the bot can fight against all weather condition. I am looking forward to see more outcome on this.
Jonathan Moon@jmoonio

For the past 12 months, I’ve been heads down building a robot. Introducing Emma – an autonomous robot that scans farms, detects diseases, and measures yield. Currently deployed in 14 vineyards and orchards in CA and NY.

English
0
0
0
18
stash
stash@stash_pomichter·
last week we got 1M views and 100s of death threats for giving Openclaw access to drones, humanoids, quadrupeds, and other physical hardware. Now we’re releasing EVERYTHING open-source. Dimensional gives agents access to the physical world. Join us. Repo 👇🏽👇🏽👇🏽
English
142
293
1.9K
151.3K
Tim Li
Tim Li@TimLi_DR·
@drfeifei @drfeifei World generation is powerful, but grounding it with real embodied interaction data will be critical. We’re building task-level data engines at DeepReach to capture real robot experiences across diverse environments. We can provide 200k hours.
English
1
0
0
55
Reflex Robotics
Reflex Robotics@ReflexRobot·
At Your Service, Episode 1: Good Morning
English
33
72
561
81K
Mark Gadala-Maria
Mark Gadala-Maria@markgadala·
So cool. Chinese AI studios are now creating full TV show series using Seedance 2.
English
267
820
5.4K
344.7K
Tim Li
Tim Li@TimLi_DR·
ABB + NVIDIA Omniverse is an important signal. But it also highlights a bigger problem: Robotics doesn’t have a hardware problem. It has a deployment problem. Most automation is built for large factories. The real opportunity is SMB manufacturing, where robots must be: lightweight affordable fast to deploy Physical AI will scale through data and deployment engines, not just better robots. #Robotics #PhysicalAI #EmbodiedAI #Automation #DeepTech
ABB Robotics@ABBRobotics

Today marks a major step for industrial automation. ABB Robotics and @nvidia have closed the sim-to-real gap - achieving 99% accuracy with RobotStudio® HyperReality. Design, test and optimize production lines virtually with unprecedented confidence.

English
0
1
1
186
Tim Li
Tim Li@TimLi_DR·
Synthetic data and lab settings can only take embodied AI so far. The real challenge for production-grade dexterity lies in the stochasticity of the field. Body:We’re addressing the domain shift by capturing high-fidelity Ego-centric (1st-person) sequences directly from the frontlines: 1. Active US Warehouses 2. Industrial Manufacturing. 3. Agricultural Operations The Research Value: Unstructured Ground Truth: Moving beyond "clean" datasets to capture real-world occlusions, variable lighting, and complex hand-object interactions. The "Last Inch" Problem: Specialized manipulation sequences designed to bridge the gap between high-level perception and fine-motor control. Long-tail Distributions: Authentic industrial edge cases that are nearly impossible to simulate accurately. Providing the foundational visual intelligence required for next-gen VLA (Vision-Language-Action) models to master task-specific physical mastery. DM to discuss our data acquisition methodology or collaboration. #EmbodiedAI #Robotics #ComputerVision #EgoCentric #VLA #FoundationModels #SimToReal #MachineLearning
Tim Li tweet media
English
0
0
0
25
Tim Li
Tim Li@TimLi_DR·
Still thinking about the Physical AI / robotics meetup we hosted in San Jose this week. We had 3× more people register than we could accommodate. Researchers, founders, and investors working on world models, robotics, and real-world AI showed up. One thing became very clear during the conversations: The biggest bottleneck for robotics right now isn’t just better models. It’s real-world deployment and data. Robots need environments to learn, tasks to execute, and systems that can actually run them in production. This is exactly the problem space we’re working on at DeepReach. Feels like the Physical AI era is just getting started. #PhysicalAI #Robotics #EmbodiedAI #WorldModels #DeepTech
Tim Li tweet mediaTim Li tweet mediaTim Li tweet mediaTim Li tweet media
English
0
1
8
10.6K
Tim Li
Tim Li@TimLi_DR·
@chelseabfinn Hi Prof Finn, how can I get contact with your team since we are building deployment OS for physical AI. I’d like to utilize your most updated model for our client production environment.
English
0
0
0
86
Chelsea Finn
Chelsea Finn@chelseabfinn·
We added short-term visual memory + long-term text memory to pi models. 🤖 Enables robots to: - complete tasks up to 15 min long - cook grilled cheese while keeping track of time - adapt in-context Paper & videos: pi.website/memory
English
9
50
490
29.9K
Tim Li
Tim Li@TimLi_DR·
Memory is powerful. Most of what makes us human comes from memory. If a bot remembers something from your last interaction and brings it up naturally, you immediately feel a sense of humanity.
Physical Intelligence@physical_int

We’ve developed a memory system for our models that provides both short-term visual memory and long-term semantic memory. Our approach allows us to train robots to perform long and complex tasks, like cleaning up a kitchen or preparing a grilled cheese sandwich from scratch 👇

English
0
0
0
33

ค้นพบ

@nvidia @DeepReach_AI @OpenAI @xAI @Meta @UCBerkeley @amazon @Apple