Simplismart AI, a burgeoning player in the artificial intelligence (AI) industry, has recently secured $7 million in funding to enhance its capabilities in machine learning operations (MLOps), focusing on streamlining AI model deployment and performance optimisation. This development addresses a critical need within enterprise settings, where organisations strive to integrate AI technologies efficiently and with cost-effectiveness.
The challenge of transitioning AI models from pilot stages to robust production environments is well-documented. It is estimated, for instance, by Peter Bendor-Samuel, CEO of Everest Group, that a staggering 90% of generative AI projects do not reach production. Similarly, research firm Gartner forecasts that a substantial portion of generative AI initiatives will be abandoned post-proof of concept by 2025. Central to these challenges is the issue of orchestration—enterprises often find themselves dependent on costly and rigid third-party APIs due to limited internal resources.
Simplismart AI aims to bridge these orchestration gaps with its comprehensive MLOps platform. This tool facilitates the entire lifecycle of AI models from fine-tuning to deployment and monitoring. The cornerstone of this innovation is a personalised software-optimized inference engine that enhances model performance—achieving an impressive throughput of 501 tokens per second on the Llama3.1 8B model—while reducing operational expenses.
Co-founded by Amritanshu Jain, formerly of Oracle, and Devansh Ghatak, an ex-Google engineer, the startup stands out in the competitive MLOps landscape thanks to its bespoke inference solutions. These solutions can manage a variety of models including text-to-speech and image processing, ensuring swift deployment and high efficiency.
The platform supports enterprises in deploying models internally to ensure privacy and control. Users can rely on Simplismart’s infrastructure or integrate their own computing resources to tailor deployments. The intuitive dashboard provided by Simplismart allows users to configure various parameters easily, ensuring a seamless orchestration of computing environments. Furthermore, the platform’s observability features enable users to monitor model performance in real time, ensuring service level agreements (SLAs) are met and performance benchmarks are sustained.
Simplismart's proprietary inference engine optimises across several layers. It enhances application serving via a custom layer for managing machine learning workloads. Infrastructure support is enhanced through rapid scaling and optimal GPU utilisation, while interaction between the model and GPU is improved using 28 custom kernels leveraging CUDA technology. These optimisations have reportedly facilitated substantial performance gains across popular models like Llama 3.1 8B, Whisper v2 by OpenAI, and SDXL.
Confronted with competition from significant entities such as TogetherAI, Baseten, Replicate, Fireworks, and Amazon Bedrock, Simplismart AI has nevertheless carved out its niche. It currently serves 30 enterprise clients, including notable names like Invideo, Dashtoon, Dubverse, and Vodex. For instance, a pharmacy marketplace has leveraged Simplismart’s platform to digitise hand-written prescriptions more efficiently, processing images 2.5 times faster at half the cost.
Looking ahead, Simplismart AI plans to further enhance its platform’s performance with the newly acquired funding. The company intends to focus on research and development with an aim to continue advancing AI inference techniques to maintain competitive superiority. In a bid to realise ambitious growth targets, Simplismart AI aims to increase its annual recurring revenue significantly over the next 15 months, targeting AI-first enterprises and promoting open-source adoption of its orchestration language.
The strategic funding and ongoing enhancements underscore Simplismart AI's commitment to addressing prevailing challenges in AI deployment and operational performance, making it an important player in the evolving AI landscape.
Source: Noah Wire Services