AI on His Own Terms: How One Computer Vision Expert Makes Ideas Happen

The possibilities driven by artificial intelligence (AI) are exploding—but the amount of talent available is not: Talent scarcity is the most significant barrier to the adoption of emerging technologies.

That skill shortage is creating a wealth of options for independent professionals like autonomous AI expert Abay Bektursun, who helps his clients explore what’s possible.

“This is my dream—to work in this technology and on my own terms,” said Abay, an Austin, Texas-based software engineer who specializes in computer vision. His clients want to know what’s possible with AI and he’s excited to be part of such rapid evolution. “I have no doubt AI will change the world for the better and improve people’s lives,” he said.

Captivated by the beauty and promise of AI

Abay graduated from the University of Central Arkansas with a degree in computer science in 2015, launching a career that quickly centered on computational neuroscience and machine learning. He developed large-scale AI projects with companies like Hewlett Packard, Eagle Eye, and Apple: “If you’ve ever been to an Apple Store, my product has probably scanned you and tried to optimize your experience. It’s all over the world.”

Abay launched a freelance business in 2022 in order to have more autonomy and more time to spend on his own projects.

“AI has been both a spiritual and technical journey for me—plus, all the opportunities make good business sense,” he said. “The main reason I got into AI was trying to understand the meaning of life. In studying AI, I hit a few nuggets of understanding of how intelligence works and it has been mind-blowing! It opened up completely new worlds of understanding of reality.”

Computer vision, Abay’s niche, is one of the fastest-growing fields in AI. It’s a subspecialty that focuses on training computers to recognize, process, and make sense of people, surroundings, and things through images, video, and other digital inputs.

While computer vision has its limitations, in specific situations it can add accuracy and reliability that go beyond what human sight is capable of—particularly with the benefit of enhancements such as infrared or magnification. Humans are flexible and adaptable, but can also be susceptible to fatigue, distraction, or “error blindness.” AI doesn’t have the same constraints. It can be trained to detect patterns or features, and interpret that data, with unmatched speed and precision. That capability powers a range of applications, from facial recognition to self-driving cars and inventory management.

As Abay explained, computer vision has two branches: discriminative and generative.

• Discriminative computer vision is the traditional approach. “When the AI looks at an image, it can recognize certain patterns or objects,” Abay explained.
• Generative computer vision is able to use existing data to generate images from scratch—and Abay said its capabilities are still to be determined: “People are still finding all kinds of new applications for it.”

Ultimately, Abay sees computer vision and other types of AI coalescing. It’s that potential that has him captivated.

“We’re realizing that, in large language models, the thing that’s being learned is not just language. LLMs are creating an abstraction of the world model,” he explained. LLMs create a map of knowledge that captures different data points, such as patterns, relationships, and concepts.

“It seems everything in nature—sounds, infrared, time, movement, language—can likely be solved by one model. It's a beautiful unification, a beautiful realization of what reality is and how you can understand it.”

Exploring what’s possible in AI

“Abay embodies the ‘think different’ mentality. He has great agency and has done an exceptional job both planning and implementing our machine learning project from start to finish. It's a challenging problem, but he's up for the challenge.”
— A client of Abay’s on Upwork

Tools like ChatGPT and Dall-E, with widespread accessibility and user-friendly design, have pushed generative AI to early majority adoption. They’ve sparked a surge of creative experimentation with countless ideas for a new generation of applications.

“I love finding people who have cool ideas but don’t quite know how to make them happen,” said Abay. “My clients may know that they can use AI but don't have the technical, research, or mathematical skills to make it happen. That's where I come in. I bring all my knowledge from research, from engineering, from all the big and small companies I’ve worked with, and I turn their ideas into reality.”

One of Abay’s client projects evolved into a startup, drawing him in as a co-founder. “I started to validate a prototype for a model in finance that would be able to understand and comprehend documents broadly.”

Over seven months, the team created a product as well as an end-to-end transformer-based model that they built and trained from scratch. Using the speed and efficiency of 200 graphics processing units (GPUs), they spent a week training the model with millions of images.

“We were the first startup that was able to—with high accuracy—extract everything from visual tables,” Abay said.

More recently, he’s consulted with a Fortune 500 manufacturing company to add computer vision to their fabrication process. “If there are defects in your product, you can visually detect it,” he explained. “A lot of factories have people who do this manually—they need to have years of experience.”

He’s working with the company to create an AI model that can discriminate certain things from images in order to automate part of the process.

How your business can approach new AI projects

“Using AI is not about gaining an edge at this point. Your business just will not exist.”

Businesses are making a big pivot to incorporate AI solutions. According to a survey by HubSpot Blog Research, 62% of business leaders said their company invests in AI and automation tools for employees to leverage at work; 43% plan to increase that investment over the coming year.

1. Understand the potential impact

For businesses that want to bring in AI, Abay would start by talking to other businesses to find out whether the solution you’re considering can actually help with your specific problem. “AI does apply to a lot of things, but it might not apply to your particular problem yet,” he said.

2. Validate the idea

Build a prototype, then see if you can make it bigger. “If you’re going to incorporate AI, you want to make sure it’s actually going to benefit you at scale,” he said. “You could find an expert on Upwork, schedule a consultation for an hour or two, and ask them a couple of questions to see if your idea makes sense. If it does, find someone to do a prototype, then keep iterating on that.”

3. Find sources of data

Abay feels more and more visual tasks will be automated as discriminative traditional machine learning and computer vision become cheaper and easier. That requires a lot of data. Rather than focusing on what the AI technologies might be, he’d put his energy into figuring out how to gather and annotate the training data you’d need as quickly as possible.

4. Prepare for unexpected change

Abay isn’t just enthusiastic about the evolution of AI-driven technologies, he’s also inspired by the optimizations AI enables elsewhere. For example, a recently released paper from Google DeepMind introduced AlphaDev, a generative model that was able to come up with a faster sorting algorithm. These algorithms underpin numerous aspects of digital society and are used billions of times a day.

“Sorting algorithms have been studied for decades—I would even say a century—and humans couldn’t see any way to improve them further. It’s been such an over-optimized field,” he said. AlphaDev has made a significant discovery—“and it probably did so in a few days,” Abay added. It underscores how researchers are constantly gaining ground, often in ways that are hard to predict.

Keeping up with the blistering pace of AI innovation

Much of Abay’s learning comes from the hands-on work of getting things done, whether it’s for clients or for himself. “I’m heavily focused on building things,” he said. “I do a lot of prototyping—even things that are not related to my clients. Like trying out new models and new research papers.”

He recently built a model that records promises that he makes: “Every time I say I will do something, it makes note of it and sets a reminder. It’s really cool.”

However, Abay also leverages the power of communities to surface the most compelling research and developments.

“Twitter is literally the best place right now because there's a character limit. AI experts are forced to summarize the golden nuggets,” he said. “If you're interested in an idea, they post the URL and you can go read the paper.” (If you’re looking for a place to start, check out @DrJimFan and @AndrewYNg.)

Abay is also one of the leaders behind Austin Deep Learning, a group he co-founded when he moved to Austin in 2016. “The group started as a way to build a network and make friends. There are so many smart people in Austin—it’s becoming a beautiful hub for AI.”

Austin Deep Learning now connects more than 3,200 members with diverse skill sets who want to consider the technology and better understand both the implications and opportunities. “We might break down the mathematics of AI or ask questions like, ‘What are the philosophical implications of this technology?’” Abay said.

Shifting your career toward AI

Generative AI is a very broad field and Abay said that to capture it completely, you’re going to need advanced education, such as a doctorate or PhD. “You have to have been doing this for decades if you want to capture everything,” he said.

However, most people won’t need to go nearly that far in order to reskill for the age of AI.

“For example, if you just want to work with LLMs—they’re extremely powerful and might actually be enough for an entire career—you only need Python skills and some critical thinking.” Python, a server-side programming language that can be used with any operating system, is one of the most popular programming languages in the world.

Then, Abay said, you can use the OpenAI API and Google to get a high-level understanding of how large language models work. “You can build amazing things that a long time ago—a few years ago, maybe even a few months ago—would have required deep computer science skills,” he said.

Plus, there’s the benefit of generative AI tools themselves. Abay said his job as a software engineer has been transformed; he’s even building in languages he’s never learned. “My productivity right now is anywhere from 2x to sometimes to 10x.”

When you first start using a tool like ChatGPT, he explained, your efficiency goes down because you need to figure out what the model is actually capable of. You’ll realize that it doesn’t always do what you want and it makes mistakes.

Eventually, however, you start to understand what makes a good prompt. “You learn: How do you verify what ChatGPT is doing? How do you work with it to make sure it does what you want?” Abay said. “Prompt engineering is a real skill. Once you have that, your productivity goes up.”

Independent experts can help chart the way forward

“I feel like there will be many stronger small teams that are building amazing products and shifting paradigms. And we'll see bigger companies move into completely new fields.”

When it comes to producing at his best, Abay knows what works for him—and it isn’t the rigid structure of a 9-to-5 office job.

“I’m more efficient and a lot freer when I can set my own schedule,” he said. “People are wildly different—that’s what makes us powerful and strong. We have different talents. Why not utilize that? Why not harvest that?”

Being an independent professional gives people like Abay room to express their strengths, define a schedule that enables them to work efficiently, and pursue the priorities that make them happiest. That, he said, is what Upwork enables.

To find an AI professional who can provide expert insights into your next initiative, contact Abay Bektursun or search for another AI and machine learning service provider on Upwork.

Disclosure: Upwork is an OpenAI partner, giving OpenAI customers and other businesses direct access to trusted expert independent professionals experienced in working with OpenAI technologies.

Upwork does not control, operate, or sponsor the other tools or services discussed in this article, which are only provided as potential options. Each reader and company should take the time to adequately analyze and determine the tools or services that would best fit their specific needs and situation.

‍