<100
Millisecond sandbox provision times
200h
Hours saved per week on sandbox infrastructure maintenance
6
Months of engineering time saved partnering with Daytona
SambaNova builds full-stack AI infrastructure powered by its Reconfigurable Dataflow Unit (RDU) chip, delivering faster, more energy-efficient inference. Since 2017, it has partnered with Meta, AWS, and Hugging Face to bring high-performance open-source AI to developers—critical for Agentic AI, as shown in their Deep Research Agents demo.
Headquarters
Palo Alto, CA
Industry
Technology
Department
Artificial Intelligence Engineering Infrastructure
Key Features
Sandbox Creation Speed Sandbox Statefulness Long‑Running Sandboxes
Learn how this full-stack, AI infrastructure solution partnered with Daytona to deliver fast and isolated execution of AI-generated code across scalable sandbox environments.
With Daytona, we now have a solution that moves at the speed of our inference.
Kwasi Ankomah
Lead AI Architect at SambaNova
01 -- CHALLENGE
Managing Docker Sandbox Environments Threatened Internal Bandwidth
SambaNova’s Agentic AI platform automates enterprise workflows, moving beyond insight generation to producing artifacts in seconds, instead of hours. However, to unlock these capabilities, SambaNova needed secure, high-speed code execution environments where agents could securely analyze data, generate outputs, and initiate action within a single trusted workflow. The challenge for SambaNova was delivering this capability without compromising security, responsiveness, or ease of integration.
Traditional infrastructure made it difficult to seamlessly bridge the gap from insight to output. Agents needed a way to execute code instantly, at scale, and in true isolation. For SambaNova, the next evolution of their Agentic AI required secure sandboxes built for these demands, ensuring every action, artifact, and automation happened securely and without friction.
Lead AI Architect, Kwasi Ankomah, observed developer momentum stalling as his team wrestled with unscalable sandbox infrastructure maintenance. And, with every new user or agent sandbox created, security vulnerabilities would multiply by the day.
Even with incremental efforts to optimize, like tuning configurations and building layered images to reduce load times, SambaNova sought more speed. They needed sandboxes to move as fast as their inference. Kwasi explains, "Our story is speed. We serve tokens really fast and really efficiently, and we want our sandboxes to be really fast and efficient, too."
The SambaNova platform required native agentic functionality, a “better together” approach where fast inference tightly integrates with an environment capable of handling many LLM calls, supporting complex workflows, and generating polished outputs, all inside secure, isolated sandboxes. Kwasi knew SambaNova needed infrastructure that could support its inference securely across thousands of agentic workloads. So, he searched for a partner who could deliver secure, scalable sandbox provisioning at the speed of inference without the ongoing infrastructure maintenance.
That’s when Kwasi discovered Daytona. The platform's agent-native infrastructure aligned perfectly with SambaNova's needs.
By the time you get to 30 Docker instances, you’re no longer building agentic applications—you’re managing infrastructure. Daytona automated that burden with instant sandbox provisioning and low-lift maintenance, so our team could get back to advancing our product.
Kwasi Ankomah
Lead AI Architect at SambaNova
02 -- SOLUTION
A Managed Runtime Platform for Seamless Sandbox Provisioning and Maintenance at Scale
After a turnkey onboarding, SambaNova quickly configured its Daytona environment while integrating its Python SDK and API with minimal setup. The team deployed their custom images via Daytona’s managed runtime, gaining instant access to hundreds of secure, isolated sandboxes that required minimal infrastructure management.
SambaNova partnered with Daytona to offload the complexity of securely orchestrating sandbox environments across thousands of users. With automated provisioning and isolation handled by Daytona, Kwasi and his team saw immediate performance gains—sandbox creation time shrunk to less than one second, and warm pools enabled agents to spin up sandboxes even faster.
With Daytona, SambaNova eliminated the overhead of maintaining compliance on self-managed Docker infrastructure. Daytona’s managed runtime enforces security and privacy at every layer, with encrypted isolation, granular access controls, and zero shared compute between tenants. Its audited HIPAA and SOC 2 certifications, combined with built-in GDPR data-handling policies, give SambaNova a verifiable compliance foundation with no internal operational load.
In addition, scalable infrastructure deployment matched the pace of every user and conversation. Kwasi recalls, “We don’t know at runtime how many sandboxes we’re going to use. Daytona allows us to auto-scale to the number of users—if 10 more users come and create a hundred more conversations, that’s all managed by Daytona.”
Beyond increased speed and scalability, SambaNova leverages Daytona to support deep research, artifact creation, and model benchmarking workflows. Engineering teams run performance evaluations such as latency and time-to-first-token reporting within secure sandboxes, while product teams use the same infrastructure for generating charts, PowerPoints, and XLS-based summaries. These rapid, sandbox-driven workflows have become integral to how SambaNova benchmarks inference performance and automates insight generation across its agentic applications.
By offloading sandbox infrastructure to Daytona, SambaNova’s team can now scale agentic workloads effortlessly—accelerating deployment cycles, expanding user access, and bringing new features from prototype to preview in record time.
One thing that Daytona does incredibly well is its sandbox provisioning times. When you're provisioning tens of thousands of sandboxes, those milliseconds add up, and no other solution we tested could match their speed.
Abhi Ingle
Chief Product & Strategy Officer at SambaNova
03 -- RESULT
SambaNova Provisions Sandboxes in Less Than 1 Second With Daytona
With Daytona, SambaNova unlocked the secure and scalable compute environments needed to match its best-in-class AI inference. Daytona’s sandboxes enable SambaNova to scale agentic workloads in real time, supporting its team of engineers as they deliver accelerated AI innovation.
<100 millisecond sandbox provision times
200 hours saved per week on sandbox infrastructure maintenance
6 months of engineering time saved partnering with Daytona
For SambaNova, Daytona is more than a fix for a technical obstacle. They’re a long-term strategic partner that enables SambaNova to overcome infrastructure headaches and keep moving at the speed of their inference.
Want to see native agentic functionality in action? Check out SambaNova's Agent Preview.
Daytona gives us a runtime platform that can execute at-speed and at-scale—the possibilities are limitless.
Abhi Ingle
Chief Product & Strategy Officer at SambaNova
