The Future of Cloud Optimization: Insights from Sync Computing’s CEO, Jeff Chou

As AI adoption accelerates, so do the challenges of managing the vast computational resources needed to power it. Cloud inefficiencies, soaring compute costs, and a growing reliance on GPUs make optimization a critical need for enterprises. 

Jeff Chou, co-founder and CEO of Sync Computing, joined Alation’s Data Radicals podcast to discuss how his company is tackling these challenges with declarative computing—a method that flips the script on infrastructure management by automating and optimizing cloud workloads in real-time.

From high-performance computing to AI optimization

Chou’s journey to Sync Computing began with a fascination for hardware-software interactions. Holding a PhD in Electrical Engineering and Computer Science from UC Berkeley and conducting postdoctoral research at MIT, he explored analog computing—a paradigm where physical systems naturally settle into optimal solutions. “What nature always wants to do is minimize energy,” Chou points out. “And if you program the physical system correctly, what it settles to can be the answer to your mathematical question.”

Though Sync no longer directly applies analog computing, its core principle of optimization underpins the company’s approach to cloud infrastructure.

“There is this gap between software and hardware… And these two worlds don't typically talk to each other… And that was the source of incredible inefficiency,” he reveals.

The problem: Cloud compute inefficiencies

Most engineers manually allocate cloud resources, often over-provisioning to avoid performance bottlenecks. The result? Massive waste in compute spend.

Chou recalled his own experience: “I used to ask for like 2,000 nodes to run my simulations and I had no idea why. I was like, whatever's max. My logic was, the more the merrier and I'm not paying for it.” This mentality prevails in enterprises today, exacerbating cloud costs.

While Databricks and Spark have made optimizations more accessible, choosing the right instance types, memory allocation, and compute resources remains a guessing game. Sync Computing is changing that.

The solution: Optimizing for outcomes

Sync Computing’s approach is based on a declarative model, where engineers define their desired outcomes (runtime, cost, performance) rather than manually selecting compute resources:

“Instead of a human having to pick the resources and pick all these configurations, that's really hard,” says Chou. “Most people don't know any of that stuff. But what people do understand is the outcome. How long did it take? How much did it cost? What was the latency? These are very understandable. These metrics are tied much more to the business I would say. 

So our whole thesis [was]: Why can't we flip the story? Why can't you declare the outcomes that you want? I want a one-hour runtime, I don't want to spend more than 20 bucks and I want this kind of performance on my job and then some magical system go figure out what hardware that is. That's the big vision of what Sync is trying to do.”

Sync’s machine learning models analyze workloads in real-time, continuously adjusting resource allocations to maximize efficiency. Unlike traditional static tuning, which requires taking workloads offline for optimization, Sync’s approach is live and adaptive—learning from production data to fine-tune infrastructure dynamically.

Applying Sync to Databricks and beyond

Currently, Sync integrates deeply with Databricks, optimizing Spark workloads for enterprises running thousands of ETL jobs daily. But the company is setting its sights beyond Spark—particularly on the growing challenge of AI inference.

“Training an AI model happens once in a while, but inference is running millions of times a day,” says Chou “That’s where the real costs add up.”

Given the staggering cost of running LLMs and other AI models, intelligent resource allocation will be crucial for scaling AI efficiently – which is a big reason NVIDIA has partnered with Sync.

The future: Specialization in compute

The future of computing, according to Chou, lies in specialized hardware. With Moore’s Law slowing, the next major breakthroughs will come from custom silicon optimized for AI, similar to how GPUs revolutionized deep learning.

“I think where the world is going is specialization,” Chou opines. “And obviously NVIDIA benefited from that massively, like GPUs, et cetera. So I see a lot of really exciting innovation in terms of, for example, if LLMs continue to dominate, what does a chip that only does LLMs look like? I think that's where the next frontier in terms of big hardware breakthroughs – it's specialization.” 

Why it matters

AI-driven companies are spending millions annually on cloud compute, often with suboptimal configurations. Sync’s vision is to automate this complexity, enabling businesses to reduce costs, improve performance, and scale AI efficiently.

With declarative computing, enterprises can focus on outcomes rather than infrastructure, marking a paradigm shift in how cloud resources are managed.

Curious to learn how a data catalog can help you optimize cloud spend? Book a demo to learn more.