Introducing VAgents

Brandon Bowsky is a consultant, business development professional, and serial entrepreneur. He is also the key person behind VAgents, a developer of AI-powered solutions that transform customer interactions by simplifying administrative processes, providing issue resolution from support content, and elevating experiences, all through the use of its signature AI agent technology.

“We specialize in generating consumer demand in numerous verticals and converting that demand into high-intent calls and leads that power industry leaders in direct-to-consumer sales. With a team of veteran experts in a number of fields, our core tenets include ensuring the best customer experience, educating our clients and affiliates, conducting good and compliant business, and prioritizing our integrity and reputation before and above dollar signs.”

How VAgents Uses Large Language Models

LLMs play an important role in VAgents’ daily tasks and operations and touch many aspects of the business. Currently, VAgents are using large language models for several applications, including: 

• Conversational AI in telephony and sales: Building agents that can handle inbound and outbound customer interactions, offering personalized recommendations and lead qualification.
• Customer service automation: Developing AI agents that can autonomously handle common inquiries, provide tier-1 support, and escalate complex cases as needed.
• Debt collection AI: Creating polite yet firm agents to handle early-stage debt collections, with a focus on compliance and empathy.
• Real-time data analysis and sentiment detection: Enabling agents to gauge customer sentiment during interactions and adjust their tone accordingly.

Prior to integrating their new servers from Puget Systems, they were facing a number of challenges, such as:

• Limited computational resources: Training LLMs required immense GPU power, and their prior infrastructure could not handle the load efficiently.
• Slow training times: Due to resource constraints, training models took an excessive amount of time, delaying project milestones.
• Scalability issues: As their projects grew in complexity, the old infrastructure couldn’t scale to handle larger datasets or more demanding AI architectures.

Brandon continued, “We were constantly spinning up new environments trying to replicate an on-prem scenario, but we were using a bunch of external services for GPUs. But the time it takes to get everything set up and ready, you can’t truly tell if it’s a viable solution for on-prem, especially when you’re trying to determine if it’s viable at scale. We discovered that iterating and testing systems in-house was significantly more efficient. That’s when we started looking.”

Enter Puget Systems

“We came across Puget Systems through peer recommendations within the AI research community. Given the positive feedback and your focus on custom hardware configurations for machine learning, we knew your solution would meet our growing needs.”

Several factors led Brandon to choose Puget Systems, which according to his team are essential for training, fine-tuning and running inference with large language models. The factors included:

• Optimized hardware: The inclusion of 4x NVIDIA RTX™ 6000 Ada Generation GPUs was a major selling point, allowing them to handle large-scale AI tasks efficiently.
• Powerful CPU and RAM: To support quad GPUs and minimize bottlenecks, an Intel Xeon™ w9-3495X with support for 112 PCIe lanes was selected as the CPU. It provided the necessary computing muscle and memory stability for training and deploying large models.
• Scalability: They needed a system that could scale with their increasing demands, and Puget Systems’ solution offered exactly what they were looking for.
• Strong reputation in AI: Puget’s track record for delivering high-performance systems tailored to machine learning workflows was a key differentiator.

“The implementation process was straightforward. Once the server arrived, we were able to integrate it into our AI development environment quickly. Our familiarity with machine learning pipelines meant we faced no major challenges in setup.”

We asked Brandon how the server’s performance was measuring up to expectations in terms of speed, reliability, and scalability. Brandon replied, “The server’s performance has exceeded our expectations in several ways. Regarding speed; training times have been cut by over 60%, allowing for quicker model iterations. On reliability, the system has been stable even under heavy loads, with no performance dips. As far as scalability is concerned, we’ve been able to scale our AI projects significantly, training models that were previously impossible due to resource constraints.”

“And all of this has dramatically improved our workflow. We can now train and deploy larger, more complex models much faster. This has enabled us to experiment with more ambitious architectures and fine-tune models in ways we previously couldn’t.”

Brandon elaborated on the performance improvements: “Training time has been reduced by approximately 65%, which has led to faster project completion and quicker deployment cycles. Cost reductions are significant; the ability to train models in-house has reduced the need for costly cloud computing services. And regarding inference speed, our real-time interactions, such as customer service responses, are now 40% faster, improving user experience.”

“One unexpected benefit, however,  has been the ability to explore multi-modal AI models—something we hadn’t considered before due to hardware limitations. Additionally, the system’s power has allowed us to handle larger datasets and perform more extensive model evaluations.”

“The server has been transformative for our AI development efforts. It has allowed us to take on more ambitious projects and deliver better results in less time. Your solution is a game-changer.”

Talk to an Expert

We specialize in building computer systems tailored for each of our customers – whether workstations, servers, storage solutions, or a combination of them all! We have found that the most effective way to ensure we meet your specific requirements is through direct communication. Our team of expert, non-commissioned consultants is adept at configuring systems that align precisely with your needs. They can refine a quote you’ve already saved or help you navigate our offerings by understanding your use case and workflow – all with no cost or obligation! Start a conversation with us today:

If you’d rather not wait, you can reach out to us via phone during our business hours.

Monday – Friday | 7am – 5pm (Pacific)

1-888-784-3872