Company: BoltzBit
CEO and Co-Founder: Dr Yichuan Zhang
Website: https://boltzbit.com/
About BoltzBit
Boltzbit is an AI research company headquartered in London, developing General Learning Intelligence (GLI). Founded in 2020 by former Google AI researcher Dr Yichuan Zhang and former Microsoft AI researcher Dr Jinli Hu, the company builds on academic research into Boltzmann machines to develop generative models capable of live learning in production.
Its models are designed to address the limitations of traditional pre-trained transformers by continuously adapting to new data and environments. Boltzbit’s technology is already in production, powering applications across financial services, data, and high-growth technology sectors.
The company is part of a new wave of research focused on moving beyond static, model-centric AI toward adaptive, context-driven intelligence, an approach referred to as AGI 2.0.
What Makes Boltzbit Different?
Many organisations using AI today are running critical operations on rented AI. When a third-party provider reprices, deprecates a model, or folds its capabilities into a competing product, there is no recourse. Boltzbit gives enterprises the infrastructure to own and train their AI so that what they build belongs to them entirely and cannot be repriced, or replicated.
Boltzbit’s models are powered by live learning, meaning they do not have a knowledge cutoff once in production. They update continuously, are trained on proprietary data, and get more accurate the longer they’re in use.
The live learning AI architecture also addresses what regulators and legal teams will eventually require: auditability. Boltzbit’s human-in-the-loop design means organisations have the full control of the model’s training data and the model’s answer for its outputs. That is not possible with frontier closed source models today.
Boltzbit is already production-tested in financial services, an environment where a 1% model error translates directly into revenue loss which is the standard its technology is built to. Its technology has been built from first principles by Cambridge and Edinburgh researchers, and is grounded in a body of published academic work.
Why It Matters
The AI field is crowded by the same variation of one theme: bigger models, better prompting, smarter retrieval.
Boltzbit is working on a different problem entirely: how to enable and democratise access to capabilities that allow users to create privately owned and highly customised models that are cost-efficient to train on proprietary data.
Whether that leads to a sustainable path to AGI or a more practical near-term payoff for users, the underlying research direction is distinctive and the early results are real.
Boltzbit is one of a small number of AI research companies in the world that have research capability to pursue fundamental architectural innovation.
Recent Work
Liquidnet, a leading institutional trading network, partnered with Boltzbit to transform its fixed income primary markets workflow. The challenge was processing unstructured bond deal data, previously a bottleneck that slowed access to new issues for traders and syndicate banks. Boltzbit deployed a bespoke live-learning model that receives, processes, and structures newly announced bond deals in real time.
98.7% model accuracy and 90% reduction in time from unstructured data receipt to bonds live in Liquidnet’s orderbook
Want to find out what TechRound can do for your business?
