Key to Generative AI Search Engines: Introducing GPUs for Embedding Model Development

Linq

AI startup Linq has adopted and is now utilizing Elice Cloud On-Demand. In this interview, we spoke with Chan-Yeol Choi, CEO of Linq, about their decision to adopt Elice Cloud On-Demand and their experience with it.

Q: Can you give us a brief introduction to Linq.

Linq is a startup specializing in market intelligence solutions powered by generative AI. We are developing Linq Alpha, an AI search service designed for hedge funds, and we offer solutions that analyze the content of stock-related documents and provide answers to queries. This enables our clients to obtain the necessary information swiftly and accurately.

👉 Learn more about Linq Alpha

Q: What made you decide to adopt GPUs?

At that time, Linq was developing the embedding model, which is the core of AI search, in-house. However, we faced significant challenges in securing the necessary resources for model development, creating an urgent need for GPUs. We reached out to Elsevier and adopted their GPU service, which significantly accelerated the speed of our model development.

*Embedding model: A fundamental model used in Retrieval Augmentation Generation (RAG) technology, crucial for addressing the hallucination problem in large language models (LLMs), one of the major issues in generative AI. RAG technology leverages the latest data not present in the LLM and internal company documents that should remain confidential, ensuring reliable results.

Q: How did you hear about Elice Cloud?

I met Elice at a founders’ meetup and learned that Elice was in the cloud business. Later, when we ran out of GPU resources, Elice Cloud came to mind, and I reached out to them. Some of the founders around me had used Elice Cloud, so I had already heard a lot of positive feedback about it.

Q: Why did you specifically choose Elice Cloud over other GPU services?

The primary reason is the speed of acquiring resources, and the second is the competitive pricing. The most important factor was having resources available when we needed them, and Elice Cloud was the most responsive in this regard. In terms of pricing, they offered very competitive terms compared to other cloud services.

Q: What are you using the Elice Cloud for? I’m also curious about what you hope to achieve by adopting the cloud.

We’ve used Elice Cloud to train our AI models, and we’ve seen some great results. For instance, Linq was ranked #1 in text search on Huggingface’s MTEB ‘Mass Text Embedding Benchmark Leaderboard’, surpassing Nvidia, Salesforce, Google, OpenAI, and others.

As Linq develops generative AI models that can provide the answers our customers need, our goal is to create high-performance search models that consistently delight our customers. We believe Elice Cloud will help us achieve this goal.

▲ MTEB Reference Image

Q: When you first started using Elice Cloud, did you find it convenient? Please tell us about any inconveniences you encountered.

I’m from an engineering background and studied computer science, so I didn’t have much trouble using it. Even when communicating with our developers, I didn’t face any difficulties or inconveniences except for the initial setup.

Q: What’s been your favorite thing about using Elice Cloud?

What impressed me the most was the customer support. We were in constant communication with Sooin Kim, CRPO, and Andy Nam, Head of AI Cloud. It was reassuring to know they understood our situation and were keen to provide the products we needed. They also kept us informed about how Elice could assist us with Linq, allowing us to anticipate and plan for resource requirements effectively.

Q: Was there anything in particular that you liked about Elice Cloud?

When launching a new service on any cloud, stability is paramount. Elice Cloud offers on-demand capabilities, making it easy to test on-demand whenever necessary and plan the next steps accordingly. This feature greatly appealed to me.

Additionally, I know they have released an ML API, and we look forward to utilizing it as we modularize in the future.

Q: How satisfied are you with Elice Cloud overall? How would you rate it out of 10?

I’d give it a 9. I contacted five GPU companies, and Elice was the fastest to respond, within two working days. We were in a rush, and Elice’s quick response enabled us to start trials immediately, allowing us to meet our desired release timeline. This prompt support contributed to our model’s success.

Q: Would you recommend Elice Cloud to other organizations, and if so, can you give us a brief reason why?

As I mentioned earlier, for developers, it’s crucial to access resources promptly when needed. I would definitely recommend Elice Cloud because of this. Additionally, their discounted structure for continuous usage creates a natural lock-in effect. In fact, I’ve already recommended it to many people around me and will continue to do so.

Q: Can you tell us a little bit about where you see Linq going in the future?

So far, Linq has developed models in various sectors such as healthcare, retail, and insurance. Starting this year, we’ll focus on hedge funds, developing numerous finance-specific models. Our goal is to provide the precise answers our customers need, striving to become the best in the world in this domain.

With Elice Cloud, Linq has been able to quickly acquire the necessary resources and achieve excellent results with its AI models. Elice Cloud offers AI startups like Linq a fast and reliable cloud environment, aiding enterprises in developing their solutions. If your organization needs a reliable GPU environment, consider Elice Cloud On-Demand as a service.

👉 Learn more about Elice Cloud On-Demand

👉 Get an Elice Cloud GPU service consultation

*Elice owns the copyright to this content, which is protected under copyright law.

*Without prior consent, secondary processing and commercial use of the content are prohibited.