Nosana Interview

Nosana dives deep into its history, the network's scalability, future plans, market competition and all things AI & blockchain.

Mango Agent #000

Share post

We managed to grab some spare time from the bustling busy bees working constantly at Nosana for some interview questions. Nosana dives deep into its history, the network's scalability, future plans, market competition and all things AI & blockchain. See what Nosana has to say as they lead accessible GPU solutions unto the Solana ecosystem.

  1. Origins: What were the key challenges and breakthroughs during the initial development of Nosana?

  2. Evolution: How has Nosana's focus shifted from CPU computations to GPU-based solutions, and what were the driving factors behind this pivot?

  3. Milestones: Can you highlight some major milestones achieved by Nosana since its inception?

At Nosana, our mission has always been to empower developers and break down barriers to innovation. We started by tackling the limitations of traditional CI/CD solutions, aiming to provide developers with more control and flexibility through a decentralized compute platform. While we successfully built this platform, we encountered challenges in gaining widespread adoption. The market largely viewed CI/CD as a "solved" problem, and developers were hesitant to switch to a new solution with fewer features than existing options.

However, this initial endeavor wasn't in vain. Developing the platform marked a significant breakthrough in creating a decentralized and flexible compute architecture, laying the foundation for what was to come.

As we witnessed the explosion of AI development, we recognized a new opportunity and a pressing need: the demand for GPUs was skyrocketing, outpacing manufacturing capabilities. This created a scarcity of these crucial resources, while many GPUs within personal devices like gaming PCs remained largely inactive. We saw a perfect alignment with our core mission of decentralization – what if we could connect these underutilized GPUs with the ever-increasing demand for AI workloads?

This realization led us to strategically shift our focus toward building a decentralized GPU grid. This grid would essentially function as a marketplace, connecting individuals with underutilized GPUs to those seeking the computational power for their AI workloads. This not only addressed the growing demand but also aligned perfectly with our vision of empowering developers and democratizing access to resources.

There have been several significant turning points in our journey. Building the initial Nosana compute platform laid the groundwork for our current work in decentralized computing. Most recently, the successful completion of Test Grid Phase 1 proved the functionality of our decentralized GPU grid, processing diverse AI workloads and gathering valuable user feedback.

This is just the beginning of our story. We are continuously innovating and adapting, and we are currently preparing for Test Grid Phase 2.

Handling Growth: As the network gains more users and nodes, what strategies are in place to ensure scalability without compromising performance?

As our network welcomes more users and nodes, ensuring scalability without compromising performance is paramount. To achieve this, we are implementing a two-pronged strategy:

Leveraging the Power of Solana:

Our network operates entirely on a decentralized stack of smart contracts (programs) on the Solana blockchain. Solana's strength lies in its ability to handle large transaction volumes efficiently and affordably. This robust foundation provides a solid base for accommodating network growth without sacrificing speed or transaction costs.

Building Secure P2P Communication Channels:

We are actively developing secure P2P channels between users posting jobs and the nodes executing them. These channels enable direct, private, and fast communication, bypassing potential blockchain bottlenecks. This approach reduces reliance on the blockchain for frequent interactions, further enhancing scalability and optimizing network performance.

By combining the scalability of Solana with the efficiency of P2P communication, we are confident in our ability to accommodate significant network growth while maintaining optimal performance. We are committed to providing a seamless and reliable experience for all users, regardless of network size.

Technological Adaptations: What technological advancements are being considered or implemented to support the growing demand for GPU resources on the Nosana network?

This is quite literally the goal of the Nosana Network: to provide a decentralized, scalable, and global solution where people can easily add their idle consumer GPUs to support the growing demand for GPU resources. We are working hard on making the node onboarding experience as easy and smooth as possible to onboard as many nodes as needed for the growing demand.

Protecting Nodes: What measures are in place to ensure the security of nodes contributing GPU power, especially against potential cyber threats?

All job workloads are being run in rootless containerized environments with a firewall in place and it is highly recommended to run Nosana Nodes in a clean environment or virtual machine. On the Solana blockchain, decentralized smart contracts are in charge of managing vaults where all transactions take place.

Data Integrity: How does Nosana guarantee the integrity and confidentiality of the data processed through its network?

By setting up a secure P2P channel between the users and the nodes, they can easily exchange private information between the two. The concept of 'job markets' allows users to decide what nodes they want to trust with running their workload. This could be any open nodes, verified secure nodes, or even your own set of nodes.

Network Vulnerabilities: Have there been any significant security challenges or incidents, and how were they addressed?

By running this in a decentralized and open/public way, we have several security challenges that we need to address. Things like private data/secrets, securing both the nodes and the users from each other, bad actors in the network, etc. are all things we need to consider. While this is an ongoing effort that is still in progress, several measures are in place or are in the works to accommodate these security challenges, like node reputation/slashing of stake, secure P2P channels, rootless containerized environments to run jobs in, trusted node selection, etc.

  1. Roadmap: What is the long-term vision for Nosana, and are there any upcoming features or expansions planned?

  2. Industry Impact: How does Nosana plan to influence the broader AI and blockchain industries in the coming years?

  3. Decentralization Goals: How does Nosana envision achieving its goal of decentralization in AI computations, and what are the expected milestones in this journey?

Our mission is to power the AI revolution with powerful computational resources. Our roadmap outlines the key stages of our growth. The journey is divided into five major release cycles named after a galaxy, symbolizing our quest to explore the vast universe of possibilities.

  1. Genesis: The birth of our platform, a stage of rigorous testing and refining.
  2. Galactica: The expansion phase, where we open up our network to the public.
  3. Triangulum: Our step towards increased interoperability, integrating with major machine learning protocols.
  4. Whirlpool: Where we broaden our horizons, supporting a diverse array of GPU types and devices
  5. Sombrero: Our commitment to cater to businesses of all sizes, with the introduction of financial and collaborative features.

At every stage, we focus on different aspects of our platform, aiming to provide a robust, scalable, and user-friendly GPU network. We eagerly anticipate the journey ahead and invite you to join us as we push the boundaries of what's possible.

# Genesis (v0.1 - H2 2023)

This is the platform's launch, which includes alpha and beta stages. It will entail bootstrapping the GPU network and replicating demand > supply > demand throughput to test the system from start to finish. It will involve testing the system, onboarding early adopters, and gathering input for future upgrades. During this phase, the emphasis will be on guaranteeing the platform's stability, dependability, and performance.

# Galactica (v1.0 - 2024)

The main grid will go live in the second phase. The CLI and SDK will be released, allowing anyone to connect to the network. In addition, a container node for consumer GPUs will be launched. The emphasis throughout this phase will be on scaling the network and engaging with a broader user base.

# Triangulum (v1.X - 2024)

The focus of this release is on integrating major machine learning protocols. The Community Connector Library, as well as official connectors for PyTorch, HuggingFace, TensorFlow, and others, will be released. During this phase, the platform's usability and compatibility with existing machine-learning tools and libraries will increase.

# Whirlpool (v1.X - 2025)

The platform's reach will be expanded even further, which will support a more diverse array of GPUs, including those from AMD, Intel, and Apple Silicon. The objective is to make Nosana the world's largest compute grid. Since it will be necessary to make sure the platform can handle a wide range of hardware and use cases, this phase will present major technical obstacles.

# Sombrero (v1.X - 2025)

Support for medium-sized and large businesses will be added in this release, as well as fiat currency ramping, billing, and teams. Because the platform will need to attract and support a broader range of users and use cases, this phase will include not only technological development but also business and community growth.

  1. AI Integration: How is Nosana innovating in the integration of AI technologies within its network?

  2. Emerging Technologies: Are there plans to incorporate other emerging technologies like quantum computing or advanced neural networks?

Nosana is building a decentralized compute grid for GPUs. These GPUs can be used in a scalable and cost-effective way for AI inference, making it both easier for AI developers to create new AI models and more accessible for companies/users that want to use these AI models. We are continuously improving our network for different types of AI inferences, like deep neural networks, large language models, etc.

  1. Community Engagement: How does Nosana plan to foster and maintain an active and engaged community?

We firmly believe in building and maintaining an active and engaged community and we do so by developing and sharing content that resonates with our audience, fostering interactive and meaningful interactions on social media platforms, and creating a feedback loop that incorporates community input into our decision-making processes. We recognize and reward our active members, organize regular community events, and offer opportunities for collaboration, as well as encourage continued engagement and loyalty. Maintaining transparent communication and a supportive environment is one of our closely held values, and we work very hard to ensure that our members feel valued, heard, and connected to our goals and have a safe and fun environment to interact with us and each other.

  1. Strategic Partnerships: What kind of partnerships is Nosana seeking, and how do these partnerships benefit the network?

  2. Collaborative Projects: Are there any notable collaborations or projects in the pipeline that the community should be excited about?

At Nosana, we believe in the power of collaboration to unlock the full potential of decentralized GPU grids. We actively seek various strategic partnerships and collaborative projects to strengthen our network and expand its reach.

As we approach Test Grid Phase 2, we are actively seeking partnerships in the AI sector, particularly with AI developers, start-ups, and institutions. This will enable us to test the platform and provide optimal solutions for this target group.

Additionally, we are looking to partner with universities and research institutes. We can provide researchers with access to powerful computational resources for their projects, while their expertise and insights can further improve our network's capabilities.

We have a few exciting developments in store that we are sure the community will be thrilled about, but we cannot share them just yet.

  1. Market Analysis: How does Nosana perceive the current and future market landscape for decentralized GPU grids?

Due to the unique features of the massive and growing GPU market, there is a lot of potential for unmet demand and underutilized supply to be matched by a decentralized and distributed marketplace.

Training large language models requires powerful GPU microchips, particularly H100s produced by Nvidia, which are 9x faster for AI training than their second fastest rival, Nvidia’s A100s. Due to various supply constraints, these are nearly all reserved for large cloud providers and incumbent tech giants, leaving an underserved long tail of AI start-ups, developers, and researchers. This leaves a multi-billion-dollar opportunity for SaaS and marketplace businesses to connect disparate idle compute capacity or more creative decentralized solutions like Nosana.

Unlike traditional cloud providers, Nosana leverages blockchain technology to create a secure and transparent marketplace where users can directly rent and share underutilized GPU resources. This peer-to-peer model eliminates expensive intermediaries and reduces costs for both providers and consumers.

The high demand for powerful GPUs, especially for tasks like deep learning, creates challenges. The need for more efficiency is urgent, with computational needs doubling every few months. However, building more data centers is tough due to limitations in expansion and pricing issues. The global chip shortage makes it even harder.

Renting GPUs from major cloud providers like AWS, GCP, and Azure is expensive due to Nvidia's licensing changes. This pushes data centers to opt for costly enterprise GPUs instead of more economical consumer options.

Overall, cost savings, increased accessibility, and the potential for disruptive innovation are what are driving the growth in the market for "GPUs as a service" and cloud computing. Decentralized networks like Nosana are key to unlocking scalable, cost-effective AI & ML advancements.

Thank you to the team at Nosana for providing time for the interview. We wait eagerly for the Public Test Grid 2 and to see Nosana gain even more attention as we get further into 2024.

Mango LogoSafer. Smarter. Faster.Get Started on Mango Today!Let's Go!

Transparency and integrity are core values in our mission to create products that empower individual financial freedom.

© Mango DAO 2023. All rights reserved.