Research Engineer @aws

Joined September 2018
354 Photos and videos
Pinned Tweet
Reachy and G1 living their best life, talking about communication… I think I found a solution to my loneliness…
1
15
685
Ship on Friday, Ship on Saturday, Ship now. We just made another Strands repository public. SOTA results with simple agents. It’s time to deploy.
I'm gonna deploy today, who's gonna stop me
1
8
404
Cagatay Cali retweeted
open agents on @aws sagemaker ai with models from the @huggingface hub via strands deployment, tool use, mcp integration, reasoning traces, and more in the blog post below
3
7
12
8,542
Cagatay Cali retweeted
🎂 Strands turned 1 today Use Python or TypesScript to build the next big AI thing!
6
19
63
4,491
Cagatay Cali retweeted
MolmoAct2 quickly becoming the default OOTB VLA baseline. Highly recommend you give it a try!
The community didn't wait for us. Before we even released code for fine-tuning, training, and inference, builders were already deploying MolmoAct2 in the wild. So we're shipping faster. Today: official LeRobot integration for MolmoAct2. Train, evaluate, and deploy with standard LeRobot datasets and workflows — bring your own task, bring your own embodiment. → github.com/allenai/molmoact2 Can't wait to see what you build.
4
13
1,464
Anyone knows better Mandarin than me?
7
731
Cagatay Cali retweeted
When we open-sourced Strands Agents SDK last year, we wanted to bring builders an easier way to build and run AI agents with just a few lines of code. Since then, the Strands Python SDK has been downloaded more than 25 million times, with more developers building with it every day. Now, TypeScript developers can enjoy the same experience, with full type safety, custom tools, and the ability to run agents in both Node.js and the browser. Strands is an agent harness SDK that goes beyond the core agent loop. It gives developers the default tools and control layers they need to build agents that can execute code, work with files, manage their own context, and stay on task as they scale. TypeScript 1.0 ships these harness capabilities built into the SDK. Default tools for shell execution, file editing, HTTP access, and structured notes, alongside customizable hooks and plugins that give developers control at every point in the agent lifecycle. All you need is a few lines of code to get started. We're building this in the open and contributions are welcome. Whether it's a bug fix, a new feature, or a cool example, we'd love to see what you build. Join us on GitHub and let us know what you think. strandsagents.com/blog/stran…
3
25
79
12,448
Android device is the best sensor array you can hand to an agent. So I gave my Pixel to my agent. But every adb/mcp tool out there is halfway done, shell works, sure. But the agent can't see the screen, can't discover UI elements, can't figure out how to unlock, can't tap semantically. No native eyes. No understanding. So I added eyes.
1
4
257
Zero config with DevDuck: DEVDUCK_TOOLS="strands_adb:adb" That's it. Your agent now has a phone. Or with any Strands agent: from strands_adb import adb agent = Agent(tools=[adb]) agent("take a screenshot and tell me what app is open")
1
130
In Turkish our elders always says something irritating for our meta cognitive development! "Komşunun çocuğu X, sınavda 100 almış." My robot is not skating dude. I need to buy some rollerblades now!
This is incredibly cool, Unitree G1 on inline skates!
14
654
Cagatay Cali retweeted
the managed agent harness in Amazon Bedrock AgentCore. define your agent, pick a model, add tools, and run it in three API calls. no orchestration code. → go.aws/4u6iSPk
3
20
1,970