dansitu Profile Banner
Daniel Situnayake Profile
Daniel Situnayake

@dansitu

Followers
5K
Following
8K
Media
25
Statuses
277

Director of Applied Research at @EdgeImpulse/@Qualcomm πŸ‘¨πŸ½β€πŸ’» Ex-@GoogleAI. Coauthor of "AI at the Edge" and "TinyML" (@OReillyMedia). πŸ‡¬πŸ‡§πŸ‡ΊπŸ‡Έ

Charleston, SC, USA
Joined January 2009
Don't wanna be here? Send us removal request.
@dansitu
Daniel Situnayake
3 years
So excited to share that _AI at the Edge_ is available for purchase! It's a practical guide for engineers, PMs, and engineering leaders, written by @jennymplunkett & I. The ebook is available right now, paper copies in a week or so. Grab yours here πŸ‘‡πŸ½πŸ‘‡πŸ½ https://t.co/naEw7WUlWi
1
56
313
@dansitu
Daniel Situnayake
3 months
This is pretty high stakes: a big chunk of our economic growth is currently based on AI capital expenditure. What happens if the race suddenly ends?
0
0
2
@dansitu
Daniel Situnayake
3 months
If model improvement plateaus, the AI industry business model is in danger. Most people will be happy to use a "good enough" model that runs locally. Why pay for a ChatGPT subscription if you have an almost-as-good model built into your MacBook/ThinkPad for free?
2
0
2
@hdevalence
henry 🌘
4 months
oh, i get it now. the reason for docker is because it's literally easier to bundle an entire operating system than a single python package
105
426
10K
@dansitu
Daniel Situnayake
4 months
TIL you can store images in a bird
@dbgray
Db
4 months
Bird saves and reproduces data: A PNG image of a bird ( photo of a bird -> spectral synthesizer ) was reproduced by an adult Starling bird https://t.co/eW4G6d4kdn. It seems to have reproduced the sound in conjunction with some additional notes which made it not detectable
0
0
0
@dansitu
Daniel Situnayake
4 months
General purpose humanoid robots are a stepping stone: they'll help us bootstrap our way to a world of individualized, task-specific robots which would not otherwise be economical to make.
0
0
2
@dansitu
Daniel Situnayake
4 months
This is a big new feature we're working on. How do you make use of VLMs on the edge while maintaining an understanding of quality?
@EdgeImpulse
Edge Impulse
4 months
New on the blog: Model cascading β€” our latest demo uses object detection to identify model cars placed in a diorama of a parking garage, triggering an onboard vision language model (VLM) to describe the car's details. Powerful potential! https://t.co/Sdx1mbNn1v
0
0
4
@dansitu
Daniel Situnayake
4 months
I've been away from Twitter for a long time but thinking of returning. Would anyone read my posts if I tweeted about embedded & embodied AI, and the tools we use to build it?
7
0
16
@particle
Particle
1 year
Edge AI has come a long way, but sensory perception is not enough. What will it take to build systems that can act? @EdgeImpulse's @dansitu has penned an insightful article that explores what's needed to bring real intelligence to the field: https://t.co/rCwWij3Jcl
1
1
2
@dansitu
Daniel Situnayake
1 year
Edge AI is maturing, and we’re getting pretty good at sensory perception. What comes next? In this post I explore what we’ll need to bring real intelligence to the field. I’d love to hear what you think πŸ‘‡ https://t.co/7DhrczdMUG
0
0
1
@EdgeImpulse
Edge Impulse
1 year
We're launching new tools for industry: β€’YOLO-Pro: Object detection that outperforms existing YOLO architectures β€’Application Behavior: Model improvements for real-world scenarios β€’Model Monitoring: Deploy and monitor models to quickly detect issues https://t.co/mfDVPFLb7X
0
2
15
@dansitu
Daniel Situnayake
1 year
I had a great time chatting with Jacob about embedded AI! Check out the podcast here πŸ‘‡πŸ½
@EdgeImpulse
Edge Impulse
1 year
Join along as Edge Impulse's Daniel Situnayake talks with Jacob Beningo about the challenges and opportunities of #AI on the edge:
0
0
5
@dansitu
Daniel Situnayake
1 year
Just a casual reminder that you can currently download AI at the Edge for free! https://t.co/ZjUFNfulOS It won't be around forever, so grab it while you can πŸ‘†πŸ½
3
9
57
@EdgeImpulse
Edge Impulse
1 year
New feature launch β€” Impulse Experiments! Test multiple ways to process your data *concurrently* to develop your ideal AI model, rather than starting from scratch each time a different result is desired. Available now to all users!
Tweet card summary image
edgeimpulse.com
Accelerate your project development cycles by experimenting with different impulses simultaneously β€” along with many other upgrades.
1
5
13
@EdgeImpulse
Edge Impulse
1 year
Training an effective edge AI model requires high quality data that is relevant to your use case. Our newest transformation block can now validate your data in minutes and create high quality datasets through the power of multi-modal LLMs. Read:
Tweet card summary image
edgeimpulse.com
Using Edge Impulse's newest transformation block, validate your data in minutes and create high quality datasets through the power of multi-modal LLMs.
1
4
10
@EdgeImpulse
Edge Impulse
1 year
Our newest integration gives you direct access to @elevenlabsio generatively created sound effects to use in your Edge Impulse projects for building robust datasets for audio-based AI models. Details and video: https://t.co/4BCbZX8NgU
Tweet card summary image
edgeimpulse.com
The latest news, releases, and features for building and deploying edge AI solutions
0
8
17
@dansitu
Daniel Situnayake
1 year
Data is the most expensive part of machine learning. Synthetic data helps reduce the cost of collecting and labelling data. Our new integrations with @OpenAI, @nvidia Omniverse, and @elevenlabsio make it easy to get started with synthetic data when training models for the edge.
@EdgeImpulse
Edge Impulse
1 year
New feature! Introducing our synthetic data generation integration, built directly into the Edge Impulse platform β€” a new and efficient way to work with LLM-generated images, audio, and voice to enhance your edge AI model creation. https://t.co/xj5UUxQrx6
0
0
4
@EdgeImpulse
Edge Impulse
1 year
Using our new feature that uses LLMs to automatically label data for building AI models, Edge Impulse engineer Jim Bruges has put together a walkthrough of using GPT-4o to classify and calculate cloud coverage from satellite imagery. Click to read more. https://t.co/eGLGsDrDFq
Tweet card summary image
edgeimpulse.com
With Edge Impulse's newest feature, visual data from a satellite can be labeled via GPT-4o both for cloud type and for coverage, generating two useful edge models from the same dataset very quickly.
0
6
14
@dansitu
Daniel Situnayake
1 year
A new post from me. What if new technology could bring us closer to an old idea? https://t.co/Lti0qrchYE
0
0
0
@EdgeImpulse
Edge Impulse
1 year
πŸš€ Edge Impulse has been featured on Metrology News for our groundbreaking FOMO-AD architecture, revolutionizing visual anomaly detection on edge devices:
Tweet card summary image
metrology.news
Edge Impulse’s FOMO-AD architecture offers accessible platform for visual anomaly detection on any edge device.
0
6
17
@EdgeImpulse
Edge Impulse
1 year
On June 18th, join @NordicTweets and Edge Impulse for an exciting dive into the world of AI/ML. Discover how to effortlessly create and deploy your own intelligent edge devices β€” explore diverse applications and possibilities. Secure your spot now: https://t.co/Vc3WYqrOG1 #EdgeAI
0
12
22