As artificial intelligence scales new heights, so does the cost of building and training powerful models. Amid this trend, London-based startup Inephany is taking a bold step forward. The company has secured £1.8 million in pre-seed funding to reshape how AI models are trained, cutting costs and improving efficiency.
The funding round was led by Amadeus Capital Partners and joined by Sure Valley Ventures, with backing from Professor Steve Young, a pioneer in AI and former Cambridge professor. The investment will support the company’s rapid growth as it builds a smarter, cleaner future for neural networks.
Inephany offers a Smarter Way to Train AI Models
Since 2012, AI compute power has doubled every 3.4 months — a rate far beyond Moore’s Law. This growth has created massive demand for training resources. For example, training GPT-4 is estimated to cost $60 to $100 million, with future models expected to surpass $1 billion in compute needs.
This model of brute-force scaling is unsustainable. It burns energy, strains infrastructure, and limits innovation to only the wealthiest organisations. Inephany is stepping in with a game-changing alternative — a platform that intelligently optimises training workflows in real-time.
The system uses adaptive optimisation to guide training more efficiently. It allows models to learn faster, using fewer resources, and improves outcomes while cutting compute bills. Early estimates suggest at least 10x cost savings compared to traditional methods.
This solution isn’t just efficient — it’s scalable and eco-conscious. The platform is designed to optimise any neural architecture, including LLMs, CNNs, and RNNs, making it useful across sectors like finance, healthcare, and autonomous driving.
A Veteran Team on a Sustainable Mission
Inephany was founded in 2024 by Dr. John Torr, a former Apple Siri engineer, along with Hami Bahraynian and Maurice von Sturm, both co-founders of conversational AI firm Wluper. The founding team has extensive experience in speech recognition, dialogue systems, and neural network design.
All three were frustrated by the rising inefficiencies in AI development. They had seen how scaling models became less about intelligence and more about computational brute force. Together, they envisioned a better way — one that makes model training smarter, not just more powerful.
The platform uses advanced sample efficiency techniques to reduce waste during training. It streamlines model iteration, accelerates development, and improves generalisation, all while reducing energy usage.
Though initially focused on training-time optimisation for LLMs, Inephany’s engine is model-agnostic. It will soon extend to inference-time as well, offering end-to-end savings throughout the AI lifecycle.
The £1.8 million raised will be invested in three key areas:
- Expanding the research and engineering team with top-tier talent
- Advancing the core optimisation engine to support more diverse and complex models
- Supporting enterprise adoption, as the first batch of commercial users come onboard
Reshaping the Future of Neural Networks
This isn’t just about cost reduction. Inephany is building a platform that democratises access to AI development, especially for smaller labs and emerging startups. As compute becomes harder to access, this technology makes development accessible, efficient, and sustainable.
CEO Dr. John Torr said: “Current approaches to training neural networks are inefficient and unsustainable. Our system tackles this problem head-on. With guidance, not brute force, we can dramatically cut training time and cost. We’re excited to work with top-tier investors and prepare for product launch later this year.”
Amelia Armour, Partner at Amadeus Capital Partners, added: “The Inephany team is solving one of AI’s biggest barriers — the rising cost of training. Their approach could cut costs dramatically and accelerate global innovation.”
Professor Steve Young, now chair and investor at Inephany, stated: “AI is spreading into fields like healthcare, weather prediction, and materials science. Efficient training is critical. Inephany’s breakthrough represents a step-change in model optimisation, and I’m thrilled to support their mission.”
