haykmartiros

To be clear, we tested a large set of batch sizes from 1 up to ~1e8. The smaller the batch size, the less competitive JAX was. The crossover point for JAX to be faster was amortizing across a batch size in the tens of thousands I believe. The ideal batch size for some of our applications is one to five. Please see symforce/benchmarks/notebooks for some of our timing.

SymForce is not trying to compute on large data-parallelism, but it could generate an inner kernel for that. In that way it complements TensorFlow nicely.

Part of the motivation for SymForce was in fact serious attempts (multiple months) by us to do similar things with TensorFlow XLA and being stymied by some of these effects. Many of the problems we deal with, and many problems in robotics in general, deal with small "tensor sizes", where dimensions of individual variables are often <10, and batch sizes aren't huge either. SymForce-generated C++ code is typically one machine instruction per scalar per "op" - compiling TF / JAX etc using methods like you mention definitely helps a ton, but it doesn't reach that level of efficiency for small problems. And computing Hessians for second-order optimization tends to blow up op counts even more.

Similarly, SymForce examines sparsity of individual matrices and jacobians at code generation time, so that operations with sparse matrices are similarly one instruction per op - at runtime there's no looping or indirection through sparse data structures or allocations of sparse matrices in the optimization loop. There's definitely a cross-over point where dimensions become large enough that TF / JAX / etc will be the tool to use - we'd love to provide some more benchmarks and comparisons here. For the first two experiments in the paper (see Tables II and III), we do compare to JAX and amortizing across a batch size of 1000, SymForce is still way ahead. We tried to be careful, but the code is available so please come beat up our numbers.

I think Casadi is solid. Primarily, I'd say Casadi is more specialized towards optimal control. To my knowledge it's much easier in SymForce to formulate complex objectives (ex: involving 3D poses, camera projections, Lie group operations), and has a more flexible code generation system that can generate structured APIs in multiple languages.

Skydio in San Mateo, CA. Flying robots are a lot easier and more fun to make in person!

Check out --> http://autonomy.skyd.io/

Contact me {hayk at skydio dot com}

Skydio Autonomy Team | SF Bay Area (remote possible) | C++ & Python

Skydio is the leading US drone company and the world leader in autonomous flight. Today our products are flown at scale in complex environments to capture incredible video, inspect critical infrastructure, and save lives of soldiers and first responders. Last year we raised a Series D for $170M at a $1B valuation led by a16z.

If you’re interested in joining a world-class engineering and research team that is defining a major emerging industry, dive in --> http://autonomy.skyd.io

We’re looking for a diverse combination of engineers who love making code for flying robots. Things we care about: performance optimization, systems infrastructure, first-principles thinking, 3D reconstruction, deep learning, human-robot interaction, sensor reliability, and UX design.

Positions: http://skydio.com/jobs

I lead this team and you can reach me at { hayk at skydio dot com }.

Skydio Autonomy | Redwood City, CA | Robotics / C++ / Python

Skydio is the leading U.S. drone company and world leader in autonomous flight. Today our products are flown at scale in complex, unknown environments by a wide range of customers to capture incredible video, inspect critical infrastructure, and save lives in emergency response scenarios. Our series D was at a $170M at a $1B valuation led by a16z.

If you’re interested in being a core member of a 30 person world-class engineering and research team that is defining the future of a major emerging industry, dive in --> http://autonomy.skyd.io

We’re looking for a diverse combination of engineers, researchers, and managers with strong SW skills and experience across complex products. We’re particularly interested in people with Unreal Engine, mobile game dev or deep learning experience.

Autonomy - Engineering Manager: https://boards.greenhouse.io/skydio/jobs/4424734003

Autonomy Products Engineer: https://boards.greenhouse.io/skydio/jobs/4303435003

Autonomy Engineer - Deep Learning: https://boards.greenhouse.io/skydio/jobs/4221006003

I lead this team and you can reach me at { hayk at skydio dot com }.

Skydio | Autonomy Engineers | Redwood City, CA | Some Remote | C++ / Python

Skydio is the leading U.S. drone company and world leader in autonomous flight. Today our products are flown at scale in complex, unknown environments by a wide range of customers to capture incredible video, inspect critical infrastructure, and save lives in emergency response scenarios. We’ve raised $170M at a $1B valuation led by a16z.

If you’re interested in being a core member of a 25 person world-class engineering and research team that is defining the future of a major emerging industry, dive in --> http://bit.ly/skydio-autonomy

We’re looking for a diverse combination of engineers, researchers, and managers with strong SW skills and experience across complex products. We’re particularly interested in people with mobile game dev or deep learning experience.

Autonomy - Engineering Manager: https://boards.greenhouse.io/skydio/jobs/4424734003

Autonomy Products Engineer: https://boards.greenhouse.io/skydio/jobs/4303435003

Autonomy Engineer: https://boards.greenhouse.io/skydio/jobs/4168122003

I lead this team and you can reach me at { hayk at skydio dot com }.

Skydio Autonomy | Redwood City, CA | Some Remote | Robotics / C++ / Python

Skydio is the leading U.S. drone company and world leader in autonomous flight. Today our products are flown at scale in complex, unknown environments by a wide range of customers to capture incredible video, inspect critical infrastructure, and save lives in emergency response scenarios. Recently raised $170M at a $1B valuation led by a16z.

If you’re interested in being a core member of a 20 person world-class engineering and research team that is defining the future of a major emerging industry, dive in --> http://bit.ly/skydio-autonomy

We combine a creative environment with production hardware to enable our engineers to iterate at incredible speeds. Our team was the first to demonstrate and ship a deep-learning based visual obstacle avoidance system. We are now revolutionizing 3D scanning with the ability for our drones to autonomously capture structures to build millimeter-scale 3D models.

We’re looking for engineers, researchers, and engineering managers with strong SW skills, mathematical fundamentals, and a drive to make robots fly themselves. Areas of interest include deep learning, computer vision, motion planning + controls, systems infrastructure, simulation, SDKs, and computational photography.

Autonomy - Engineering Manager: https://boards.greenhouse.io/skydio/jobs/4424734003

Autonomy Engineer: https://boards.greenhouse.io/skydio/jobs/4168122003

Research Intern: https://boards.greenhouse.io/skydio/jobs/4255866003

Software Eng - Commercial Products: https://boards.greenhouse.io/skydio/jobs/4303435003

I lead this team and you can reach out directly to { hayk at skydio dot com }.

Skydio Autonomy | Redwood City, CA | Some Remote | Robotics / C++ / Python

Skydio is the top US drone company (fundraise at 1B announced today!) and world leader in autonomous flight. Our drones are flown at scale in complex, unknown environments to capture incredible video, automate dangerous inspections, and save lives in emergencies.

Visual walkthrough of our team: http://bit.ly/skydio-autonomy

Customer highlights reel: https://youtu.be/4rdrtEU-9Bw

3D model: https://skfb.ly/6Zpvz

The autonomy team is 20 people who drive the state-of-the-art in computer vision, motion planning, and high-performance systems infrastructure to make our robots intelligent. Our drones use 360 vision and all processing is done onboard. We combine a creative research environment with real production hardware to enable our engineers to iterate at incredible speeds.

Our team was the first to demonstrate and ship a deep-learning based visual obstacle avoidance system. We are now revolutionizing 3D scanning with the ability for our drones to autonomously capture structures to build millimeter-scale 3D models.

We’re looking for engineers and researchers with strong SW skills combined with strong mathematical thinking, and a drive to make flying robots get sh*t done.

We’re also looking for world-class technical leaders in the areas of simulation, SDKs, systems infrastructure, object detection, computational photography and to lead our public sector work.

I lead this team, reach out to { hayk at skydio dot com }.

Skydio Autonomy | Redwood City, CA | Remote Possible | C++ / Python / Robotics

Skydio is the top U.S. drone manufacturer and world leader in autonomous flight. Founded in 2014, raised $170M. Our drones are flown at scale in complex, unknown environments to capture incredible video, automate dangerous inspections, and save lives.

Visual walkthrough of our team: http://bit.ly/skydio-autonomy

2020 customer highlights reel: https://youtu.be/4rdrtEU-9Bw

The autonomy team is ~20 people who drive the state-of-the-art in computer vision, motion planning, 3D reconstruction, and high-performance systems infrastructure to make our robots intelligent. Our drones have 360 vision and all decisions are made onboard. We combine a creative research environment with real production hardware to enable our engineers to iterate at incredible speeds.

For example, we were the first to demonstrate a deep-learning based visual obstacle avoidance system and shipped it on the Skydio 2, where it remains the industry best. We are now revolutionizing 3D capture to be able to autonomously map anything and make 3D models in incredible detail.

We’re looking for engineers and researchers with strong SW development skills combined with strong mathematical thinking, and a drive to work on robotic systems. We are open to an experience range from grad students to world-class technical leaders.

I lead this team, reach out to { hayk at skydio dot com }.

We are open to it for very exceptional candidates, given increased collaboration challenges

Skydio Autonomy | Redwood City, CA | Remote Possible | C++ / Python / Robotics | https://bit.ly/skydio-autonomy

Skydio is the leading U.S. drone manufacturer and world leader in autonomous flight. Founded in 2014, raised $170M. Our drones are flown at scale in complex, unknown environments every day to capture incredible video, automate dangerous inspections, and save lives.

The autonomy team is ~20 people who drive the state-of-the-art in computer vision, motion planning, 3D reconstruction, and high-performance systems infrastructure to make our robots intelligent. We combine a creative research environment with real production hardware, to enable our engineers to iterate at incredible speeds.

For example, we were the first to demonstrate a deep-learning based visual obstacle avoidance system and shipped it on the Skydio 2, where it remains the industry best. We are now revolutionizing 3D capture to be able to autonomously map anything and make 3D models in incredible detail.

We’re looking for engineers and researchers with strong SW development skills combined with strong mathematical thinking, and a drive to work on robotic systems. We are open to an experience range from new grads to world-class technical leaders.

Visual walkthrough of our team: http://bit.ly/skydio-autonomy

2020 customer highlights reel: https://youtu.be/4rdrtEU-9Bw

I lead this team, reach out directly { hayk at skydio dot com }

Love this! Ordered for the office.

For what is worth kragan is entirely right here.

The article claims the scalar exponential solution is O(1) - this is incorrect because the exponential of a scalar is still O(logN).

Skydio | Software Engineer - Skills SDK | Redwood City, CA | ONSITE VISA | Full-time

Skydio makes a flying robot platform that’s built on top of the most advanced autonomy stack available in the world today. In the hands of knowledgeable professionals across multiple industries, Skydio has an unprecedented opportunity to put autonomous UAVs to work at enterprise scale. We’re looking for a software engineer to design and build public-facing APIs that drive the growth and development of the Skydio Autonomy Platform, from early beta into product maturity and widespread adoption. Funding 70M+.

Responsibilities

* Define and lead technical development APIs for autonomous drones targeted towards enterprise and consumer markets.

* Understand high-level industry vertical requirements and create actionable API functionality to address high-value requirements.

* Develop and maintain a rigorous test and regression framework that ensure continued API safety as the product evolves.

* Develop and maintain enterprise integrations with a keen eye for security and sensitivity to intellectual property.

Requirements

* High proficiency in Python, Javascript, C++

* Experience designing, building, and managing substantial software APIs

* Strong communication skills, interface effectively with internal engineering team and external partners.

* Experience with and understanding of full-stack software principles; Security protocols, performance analysis, cloud compute resource management, documentation

* Strong analytical skills; Vector Math, 3D Rotations, Basic Linear Algebra and Calculus.

* Passion for Autonomous Flight

* Chill

Recommended

* Extensive knowledge and experience with robotics systems, specifically with Perception/Motion Planning/Controls.

* Past experience with using UAV in an enterprise or industrial setting.

* Some experience with IOS/Android development.

* Some experience with TensorFlow.

* UAV pilot

recruiting@skydio.com