Aaryan Agrawal

52 posts

Aaryan Agrawal banner
Aaryan Agrawal

Aaryan Agrawal

@aboutaaryan

helping makers of robots and machines scale field support @far__hand | author & engineer from dartmouth

San Francisco Katılım Mart 2016
71 Takip Edilen55 Takipçiler
Aaryan Agrawal retweetledi
Kyle Vedder
Kyle Vedder@KyleVedder·
if a robot deployment co isn’t obsessed with - mean time to failure - reducing part count - design for repairability - maintenance strategies it’s over for them models are the current bottleneck, but they get better at frontier lab speed — hardware + ops don’t
English
7
16
227
42.9K
Aaryan Agrawal
Aaryan Agrawal@aboutaaryan·
Good models (pi, gen) Good middleware (ros 2, viam) Good dev tools (claude code) Good hardware (unitree, realman) Who do you compete with?
English
2
0
2
48
Grateful Dad
Grateful Dad@dawghater23·
@aboutaaryan I was an integrator / OEM for ~7 years. Now (at least for the next couple days) I'm with an OEM start-up.
English
1
0
0
10
Zane Hengsperger
Zane Hengsperger@zanehengsperger·
If I could fund a single company that doesn’t exist right now it would be… Machine Maintenance as a Service $222B/year in the US is spent on machine maintenance or repair. Features would include: - quarterly PMs on machines - full service history with access to service history of other exact machines - predictive from sensors or cameras - instant buy spare parts from online store - operator training module - access to AI chat bot (ik cringe) that knows your machine inside out This would be an incredible bet as robotics scale and maintenance is inevitable. I’d hire the oldest guys with the most amount of tribal knowledge. Every single repair or process would be documented with data points and feed into our system correlated to a specific machine.
English
152
81
1.5K
378.1K
Aaryan Agrawal
Aaryan Agrawal@aboutaaryan·
if youre not using this, youre dusted
Kiet@FlyaKiet

Superset 2.0 is in open beta and it's a big rewrite. Why and what's different about this? I walked through a tour of the changes here: 0:00 - Introductions 1:53 - Why we rewrote Superset 4:00 - New file system 5:00 Automations 5:44 CLI 7:48 Thank you to early users! tldr: Superset 2.0 is in Open Beta! It's a full rewrite of the Superset app to enable Cloud and remote workspaces to work like realtime. This is the next step on the journey to enable developers to run 100s and even 1000s of coding agents in parallel. Work on remote workspaces in real-time as if it was on your local machine. Also, new simplified UI, and a a sick CLI to let agents drive your Superset app. Mobile working internally, and more coming soon!

English
3
2
8
3.4K
Aaryan Agrawal retweetledi
Kiet
Kiet@FlyaKiet·
Superset 2.0 is in open beta and it's a big rewrite. Why and what's different about this? I walked through a tour of the changes here: 0:00 - Introductions 1:53 - Why we rewrote Superset 4:00 - New file system 5:00 Automations 5:44 CLI 7:48 Thank you to early users! tldr: Superset 2.0 is in Open Beta! It's a full rewrite of the Superset app to enable Cloud and remote workspaces to work like realtime. This is the next step on the journey to enable developers to run 100s and even 1000s of coding agents in parallel. Work on remote workspaces in real-time as if it was on your local machine. Also, new simplified UI, and a a sick CLI to let agents drive your Superset app. Mobile working internally, and more coming soon!
Kiet@FlyaKiet

@superset_sh 2.0 is now in open beta! For the past few weeks, we've been hitting physical limits on using Superset, my computer begins meltdown when running 20+ worktrees. Many of our users have 50+ (!!!) Since we're 3 ex-CTOs, the logical next step is to rewrite the whole app to enable cloud workspaces. And today that's available to the public 🎉

English
12
5
91
19K
Kyle Vedder
Kyle Vedder@KyleVedder·
why are no data factories selling post-training data? take a base policy, pick a task, dagger it to three 9s reliability, and sell the dagger data as a premium bundle pretraining data is a commodity slop game, post-training is value add but no one’s doing it yet 🤡
English
43
10
285
28.8K
Nicolas Keller
Nicolas Keller@Nicolas_Keller·
We are also doing it every day with some vanilla policies for data quality checks, since we want to understand the relationship between data and policy output/outcome as much as possible. Are you even interested in that type of data, i.e. where the base policy is a vanilla flow-matching type policy instead of your own policy? We can do both.
English
1
0
1
178
Matthew Amick
Matthew Amick@mtamick·
@aboutaaryan @snowmaker Sheet metal fabrication for PACCAR. Lots of Komatsu presses, Amada lasers & brakes, and welders. I ran the maintenance department for a few years and then took over as plant manager at the powder coating sister plant down the road.
English
1
0
0
16
Loren
Loren@lomcflurry_·
Quick trip to NY tomorrow who should I meet with?
English
3
0
1
300
Ilir Aliu
Ilir Aliu@IlirAliu_·
Everyone in robotics knows this pain
Zane Hengsperger@zanehengsperger

If I could fund a single company that doesn’t exist right now it would be… Machine Maintenance as a Service $222B/year in the US is spent on machine maintenance or repair. Features would include: - quarterly PMs on machines - full service history with access to service history of other exact machines - predictive from sensors or cameras - instant buy spare parts from online store - operator training module - access to AI chat bot (ik cringe) that knows your machine inside out This would be an incredible bet as robotics scale and maintenance is inevitable. I’d hire the oldest guys with the most amount of tribal knowledge. Every single repair or process would be documented with data points and feed into our system correlated to a specific machine.

English
7
4
87
27.6K
Aaryan Agrawal
Aaryan Agrawal@aboutaaryan·
@mtamick @snowmaker about 2000 nationwide robots, automation systems, and computerized machinery. except for very large/complex ones from each of these what kind of mfg plant did you manage?
English
1
0
0
47
Kevin Bing
Kevin Bing@KevinRBing·
@zanehengsperger #1 on your list machine maintenance seems like one of the few applications where AR is additive. 6 shop floor techs supervised by 2 remote engineers. It kind of works this way now but it’s typically the senior tech standing on my loading dock on his cell coaching the others
English
1
0
3
517
Brennan M. Woodruff
Brennan M. Woodruff@BrennanWoodruff·
@zanehengsperger Alright, I’ll build an accelerator around this. Who should we get involved from existing industry players?
English
3
0
5
627
Aaryan Agrawal
Aaryan Agrawal@aboutaaryan·
@m0rcutt we are building point 1. @far__hand 2000 service engineers nationwide + internal AI tools that act as a manager + senior technician could I ask you a couple questions?
English
0
0
0
11
Michael Orcutt
Michael Orcutt@m0rcutt·
These are great. On point 1, makes a lot of sense. Ran several marketplaces and can see this working. Add an MCP to the stack and your AI shop manager now can keep track of maintenance and automatically schedule. On point 6, I recently had a small shop (~10 people) do a $5k welding job for me. I had over 50+ emails/texts back and forth. Most of it was "where are we at on this?" and then confirming each detail as they did the work. I created drawings, docs, and thoroughly outlined this work like it was a software project. The shop owner was a great guy but the experience was terrible. This extends into any "custom" work. I ordered custom cushions, laminate (from a big vendor), and other things and all were running their businesses like I had to manage them. The big vendors software was at least a few decades old and had no easy way to order online.
English
2
0
4
911
Aaryan Agrawal
Aaryan Agrawal@aboutaaryan·
@ayingerbrau88 @zanehengsperger i was a machinist. love that you're building this! we have install and repair machinery like yours across the US. 2000 service engineers + internal AI tools that help you sell more machines without worrying about field personnel. lets talk?
English
0
0
1
26
Charles
Charles@ayingerbrau88·
Love this sentiment. We are a startup building Safer abrasive saws for cutting difficult to machine materials that can’t efficiently be cut any other way. We have reduced lead times from 20 weeks during COVID down to 13 weeks and we are working on shorter. Soon we will expand to adjacent machines like grinders and other necessary equipment to supply American aerospace foundries and beyond.
English
2
0
4
777
Aaryan Agrawal
Aaryan Agrawal@aboutaaryan·
@AcoineY we are building what zane described. could i ask you a couple questions abt ur experience at Siemens? trying to democratize PLC
English
1
0
0
20
Açoine
Açoine@AcoineY·
👀👀 Part IoT, part DB, part AI. I believe my time at Siemens they were working on this stuff. Making it commercial would be great. An easy modulable monitoring service with actions you could auto set. Biggest hurdle is the physical maintenance of the machines- on staff or contracted
English
1
0
0
97
Aaryan Agrawal
Aaryan Agrawal@aboutaaryan·
@dawghater23 we are ex-roboticists working on service for all kinds of robots. chinese, taiwanese and japenese robotic OEMs have shown the most interest in the solution. are you an integrator deploying new systems?
English
1
0
0
42
Grateful Dad
Grateful Dad@dawghater23·
I work in the "new business" side of this space. The lifecycle service & support side is much more saturated than you think. Just not popular on here or other social platforms. I do know of some buddies from the larger "incumbents" in our space that are spinning up smaller companies to take LSS business from customers. These guys are all multi decade tenured DUDES from top system integrators. It's out there, but it's a quintessential good ole boy network, and you better know your stuff. As the market is shifting towards more robotics and flexible automation, the demand for state-side support is also rapidly expanding. I talk to a lot of customers fed up with calling China.
English
1
0
2
596
Pushkar Limaye
Pushkar Limaye@pushkar_24·
This only works if you start with operators, not the sensors. Captured 40 yrs of failure intuition from one operator - now queryable on WhatsApp by the next shift. Breakdown prediction sits on top of that layer, not the other way around. Shipped in weeks at a wheel rim plant in India. US May 5–8 (MIT), then Ohio/Detroit/Texas/SF - Looking forward to meeting folks deep in plant ops.
English
1
0
1
247

Keşfet

@dawghater23 @Nicolas_Keller @KyleVedder @mtamick @snowmaker @lomcflurry_ @samanfarid @IlirAliu_