Innovation in the Developer Community

Sparking AI Innovation in the Developer Community

October 29, 2024

Christina Cardoza

A trio of developers at an industry event listening to a presentation.

Businesses use AI to transform their industries, from manufacturing to medicine to education; that’s clear to even the casual observer. But behind all that transformation are developers building solutions and driving innovation. This evolution happens so fast that even professionals and experts have a hard time keeping up. Fortunately, no one needs to master the field of AI development alone; there’s a whole ecosystem of communities and partnerships out there to be leveraged.

Partnering for this discussion are representatives from two companies embedded in the AI and developer communities: Paula Ramos, AI Evangelist from Intel; and Jason Corso, Co-Founder of Voxel51. Jason is also a faculty member in the department of Robotics and EECS at the University of Michigan. They talk about where AI is going and how developers contribute to its advancement, as well as the importance of engaging in the development community—for the benefit of businesses and industry as well as for developers themselves (Video 1).

Video 1. Industry thought leaders from Intel and Voxel51 discuss the importance of fostering engagement among the developer community. (Source: insight.tech)

How is the AI space evolving and what trends are shaping it?

Jason Corso: In the past few years there have been a few major developments driving the way we all think about AI. The first one is the availability of these large language models that can capture huge token lengths and embed natural human language into the model. That will give us a resource in which we can really interact naturally, too.

We are also seeing language combined with vision as a key future trend. This will come with new compute capabilities, openly available data, and the ability to use these foundational models to really tackle new problems in what at Voxel we call “visual AI.”

Another thing I’d point to is an appreciation for the role that data is playing in the development of various AI/ML models. We’ve built this culture where the model is king. When you take a machine learning course in school, you start out training models with some data set you’ve downloaded or that the professor has given out—most of the focus is on the algorithm. But various leaders in the LLM space have begun to talk about the critical role that data, good data, high-quality data plays in this marriage of model, code, and data to build the AI systems that we’re using.

At Voxel, for example, we focus heavily on the role that data plays and on providing developer tools for engaging with data alongside the models—rather than just expecting a developer to generate up some scripts to visualize that data. Twenty years ago my data sets were dozens of samples or hundreds of samples, right? Now we have data sets that are dozens of millions of samples. So actually managing them and understanding the failure modes and the distribution and so on is very difficult and requires, I believe, new thinking.

What role do developers play in the AI advancements?

Paula Ramos: Developers are looking for their path every day because things are changing so fast. They need to drive innovation in the huge field that is artificial intelligence, so they need to be creative in order to solve problems. Maybe we have the same problems that we had 20 years ago, but there are better tools right now; there are better ways to approach the solutions. We also need to think more about the final user of the application.

“#Developers are looking for their path every day because things are changing so fast. They need to drive innovation in the huge field that is #ArtificialIntelligence.” – Paula Ramos, @Intel via @insightdottech

There are some challenges right now, and I think we still have room to grow in model development, data management, and how we can deploy those models in an easy way. Do you use a cloud system, or do you use an edge solution? The solution always has to be the simplest one possible, and this is the main challenge developers have right now.

Also, something that is really important in this field is the open source community; this is changing the cadence of the AI. When we have these models open to everyone, they can access the data sets and improve those models round by round.

What are the best ways for developers to partner with companies like Intel?

Paula Ramos: We have multiple channels right now and a variety of solutions. For instance, we have hardware accelerators for retraining or for fine-tuning models. We also have solutions that work at the edge.

There are the Edge Reference Kits that developers can access. It’s one way to give a complex problem an easy solution. And there we are trying to show them with tutorials, code, and videos how they can navigate specific verticals: manufacturing, retail, healthcare. Also for LLMs and how to work with multiple models.

Or developers can use OpenVINO^™ to optimize and quantize a model. That means that they can use the same infrastructure that they have—we are not forcing developers to buy specific hardware to run models—and they can optimize and quantize LLMs. OpenVINO also enables developers to easily prove and test these LLMs. They can create pilots and provide examples before moving to the real or final production system.

We have an amazing repository and open source community where developers can test the latest AI trends. If something new came out today, literally in two days that specific model would show up in the OpenVINO notebook repository. You can test there, for example, Llama 3.1, YOLOv10, and the latest AI trends. This is a great tool.

Developers can also access Intel Developer Cloud to test multiple kinds of hardware before buying it. That is really cool. They can access accelerators and the latest AI trends, for example, the AI PC.

How is Voxel51 engaging with developers?

Jason Corso: Our software is called FiftyOne. It’s basically a visual component as well as a software SDK for doing the work that we’re talking about here, like data and model refinement. But most recently we have this new functionality called Panels. With Panels you can build functionality for the front end without knowing how to write React or JavaScript or anything with UX. You can write it directly in Python and still enhance the GUI functionality.

As a company we are open source driven, but we do actually have dozens of customers that use our commercial-enterprise version of the FiftyOne software—called FiftyOne Teams—and it allows you to develop the same functionality together in teams, in the cloud, or on-prem. We have a pretty broad customer base across manufacturing, security, and automotive.

We closed our Series B funding earlier this year, and actually we’re hiring for machine learning engineer roles, among others—both for core engineering work as well as developer-relations work. We believe in developers so much, so we hire individuals who are fully trained and can write papers and code and so on, but their role is actually building bridges with the community.

How do industry events help developers engage with the wider community?

Jason Corso: Voxel had its first in-person hackathon just before CVPR. It’s one type of engagement where we see developers excited to be engaging with new technology and really trying to work together on new teams to solve a new problem.

That was fun, but I think a key angle for developer events is obviously education. One has to go to developer events or conferences like CVPR to really stay abreast of what’s happening. Last year I taught the course Intro to Computer Vision: In some sense, for three hours a week I was doing this developer event for 300 students to learn about computer vision.

But the AI space is evolving so rapidly that it seems like everyone is in constant information-gathering mode—even faculty members who have been in the field for ages. It’s impossible to stay up to date with everything from cutting-edge research papers all the way to the new APIs and libraries that you have to learn.

So at Voxel what we’ve tried to do is maintain a sort of weekly technical event that really allows the community to stay engaged. Just personally, for example, every Monday at noon Eastern I have open office hours and anyone can sign into them on Zoom. A couple of weeks ago we reviewed someone’s paper, and we went through slides and an actual technical model. But it goes all the way to: “This is my first time thinking about getting into computer vision. What should I be looking at?”

So there’s raw education just about foundational capabilities, but also developer events that really help engagement while staying up to date with what’s happening.

Are there any other available resources that developers should take advantage of?

Jason Corso: As Paula said earlier, being open source is the gateway to fostering innovation. Our software, FiftyOne, is on GitHub, and you can fork it and you can submit PRs. We make releases on the order of every month or two, and every release has some content from our community. We’ve been educated so much about community needs and by community contributions over the past four years since we released it. I just really want to express my thanks to the developer community that we’ve built. It’s such a vibrant and rich environment, and we wouldn’t be where we are today without that community.

There are actual events, but just becoming a part of open source projects is another great way to really get involved in the developer ecosystem for AI.

How does Intel foster community engagement for developers?

Paula Ramos: At Intel, we have been working so hard on that part, and we are creating multiple ways to create this innovation with developers. We have a huge ecosystem where we are trying to touch not just the inference part but also the training part, with anomaly detection, for example.

We have one program called the Innovator Program, where we have multiple developers around the globe testing technology. They can make their own applications and then share them with us. Basically, they create their own repository, then they fork their repository and create new applications. I will be highlighting some of these innovators in my LinkedIn and on my network, so stay tuned.

Another thing that we participate in is the Google Summer of Code, and there we have several developers working with us for three months with different mentors from the OpenVINO team. Last year, one of the students involved with the Google Summer of Code created a paper with their mentors about Anomalib. The paper was submitted to the Visual Anomaly Inspection Workshop at CVPR, and it was accepted.

We are also moving fast in relations with universities, for sure, helping them to create research and research proposals that Intel can support. We are also closing the gap in between industry and academia with conferences.

We have a real intention to work in the open source community, because here the most important thing is developers. Always we are thinking that we need to enable developers to use this hardware in software that we can provide, and developers can accelerate so that they can improve their pipelines and their workloads. That is the main intention.