While generative AI tools have automated workflows for developers and creatives, AI engineers still spend months manually fine-tuning models. This process is GPU-heavy, expensive, and time-consuming, creating a major roadblock for startups and enterprises. Inference optimisation, making trained AI models run efficiently on various hardware, remains largely manual, with up to 70% of AI system deployment costs tied to GPU infrastructure. TheStage AI aims to eliminate this bottleneck by automating optimisation, slashing time and cost without sacrificing performance.
The US-based startup has secured $4.5 million in seed funding to address a critical bottleneck in artificial intelligence: the inefficient and expensive process of optimising inference. The round brought together a group of strategic investors, including Mehreen Malik, Dominic Williams (DFINITY), Atlantic Labs (SoundCloud), Nick Davidov (DVC), and AAL VC. The funding marks a strong vote of confidence in TheStage AI’s vision and capabilities, positioning it for rapid technological and commercial growth.
Fueling the future: How does it plan to use the funding?
TheStage AI plans to utilise the funding to enhance its proprietary optimisation system ANNA (Automatic NNs Analyzer), expand its Model Library, and scale its deployment infrastructure. Investment will also go toward growing its team and deepening integrations with leading cloud platforms such as AWS, Google Cloud, and Microsoft Azure. Additionally, part of the funds will be directed at customer acquisition efforts, especially targeting app developers and model builders.
Discussing with TFN, TheStage AI stated: “In the next 3-5 years, we aim to be integrated into all major AI application development tools and AI cloud platforms. With our product, anyone will be able to create neural networks in minutes, tailored to specific tasks, datasets, and devices, including edge hardware. Users will have confidence that they’re achieving optimal performance in terms of both quality and speed.”
Brainchild of deeptech veterans with a Huawei legacy
TheStage AI was founded by four university friends – Kirill Solodskih, Azim Kurbanov, Ruslan Aydarkhanov, and Max Petriev, each holding PhDs in mathematics or neuroscience. The team brings over a decade of experience in optimising deep neural networks. Before launching TheStage AI, the founders worked together at Huawei, where they developed model compression and acceleration technologies for flagship smartphones like the P50 and P60. Their research legacy includes over 15 published papers and more than 10 patents, including work presented at top conferences like CVPR and ECCV.
The hidden cost of AI: Tackling the inference optimisation crisis
Talking about what led to this idea, TheStage AI stated: While working on AI quantisation frameworks for Huawei’s P50 and P60, members of the founding team saw firsthand how rigid and manual the optimisation process was. They built internal tools to automate quantisation, reducing development time from over a year to just a few months without compromising model quality. One of the patented algorithms developed during that time became crucial when Huawei had to shift from Kirin to Qualcomm processors due to sanctions, enabling rapid adaptation of neural networks to new architectures.
This experience demonstrated how impactful automation could be in AI optimisation, especially when speed, efficiency, and flexibility are critical. The challenges faced at Huawei reflected broader industry pain points. TheStage AI was founded to solve those problems at scale. With its core product, ANNA, the team aims to bring fully automatic, production-grade optimisation to AI developers everywhere.
What does TheStage AI do? Why does it matter?
TheStageAI further stated: “Its flagship technology, ANNA, uses discrete math and AI to automatically optimise PyTorch models through advanced techniques such as quantisation, pruning, and sparsification. ANNA can generate “Elastic models,” which vary in size and performance, giving users the flexibility to choose a model best suited to their specific hardware and latency needs, similar to selecting a video quality on YouTube. These models are available in a Model Library that includes open-source solutions like Stable Diffusion, optimised for various performance and cost scenarios. TheStage AI also offers automatic acceleration services for companies customising their own models. Importantly, the platform supports a wide range of environments, from smartphones and on-prem GPUs to all major cloud providers.”
In collaboration with Recraft.ai, TheStage AI doubled performance and reduced processing time by 20% compared to PyTorch’s compiler. Unlike competitors that lock users into proprietary hardware, TheStage AI offers flexibility in supporting a wide range of hardware setups, from smartphones and custom on-premises GPUs to cloud environments. For cloud integration, TheStage AI works seamlessly with AWS, Google Cloud, and Microsoft Azure, utilising their fixed hardware setups. The company targets two primary user groups: app developers seeking pre-optimised models for seamless integration and model builders requiring granular control for custom neural networks. This dual approach offers a significant competitive advantage over rivals that only provide ready-made solutions, effectively addressing the scaling challenges faced by startups.
Talking about the competition, the company further stated: “We are targeting the inference market, and our competitors include inference providers and hardware companies that produce acceleration tools. Key players include Replicate, Fireworks, and AWS SageMaker, which provide serverless model inference on their hardware. In the tools space, we compete with TensorRT Model Optimizer and Intel Neural Compressor. However, we’re targeting a specific niche, as we believe the inference market will expand and our automation tools will benefit these companies as well.”
Final thoughts
As AI adoption accelerates, the demand for smarter and more efficient inference is becoming more urgent than ever. Meta’s $65B infrastructure plan and McKinsey’s finding that 70% of AI deployment costs stem from GPU usage only reinforce the need for solutions like TheStage AI. By automating what was once a tedious, manual process, TheStage AI not only makes AI deployment faster and cheaper, but also opens the door for broader adoption. With its experienced team, strong investor backing, and cutting-edge technology, TheStage AI is well-positioned to become a foundational player in the AI infrastructure ecosystem.
“AI models excel at implementing ideas and logic that are difficult to express through traditional deterministic algorithms. When we deploy these AI models, we’re bringing these ideas to life. We’ve created a service that allows AI engineers and developers to compress, package, and deploy models to any device as easily as copy and paste.” said Kirill Solodskih, CEO and Co-Founder of TheStage AI.
“Investing in TheStage AI at this early stage is a great opportunity. I understand that successful solutions always combine hardware with the right software. I am helping the team with business strategy, and we plan to demonstrate significant product growth in the next several months.” added Mehreen Malik, Lead Investor in TheStage AI.
The post TheStage AI bags $4.5M to make AI optimisation effortless: Can Huawei alumni bring a difference? appeared first on Tech Funding News.