We started Alacene after years of watching teams in regulated, technical and knowledge-heavy environments try to bolt AI onto systems it was never designed to respect. We think the right answer is infrastructure, not another wrapper.
Alacene is built by a founding team with backgrounds in applied AI, distributed retrieval and information products for high-trust environments.
Scientist and engineer with publications in Nature and collaborations with a Physics Nobel laureate. Former CERN engineer and LAMBDA researcher (HSE), with experience spanning materials science and principal data science, delivering Bayesian systems for live entertainment and pharmaceuticals. Specializes in interpretable ML for physical and high-stakes systems.
PhD researcher at MBZUAI and former Senior NLP researcher in industry. Published at A* conferences on retrieval-augmented generation, uncertainty in LLMs and trustworthy QA. Shipped a factual question-answering skill inside a commercial virtual assistant used at scale.
Core developer of PyMC and active NixOS contributor. Built large-scale training infrastructure for computer vision and led Bayesian engineering across pharmaceuticals, consumer goods and live streaming. Invited lecturer in Practical Bayes at Moscow State University.
Abdalaziz works across research, modelling, infrastructure and inference at Alacene: interpretable models, uncertainty-aware inference, and tightly integrated hardware–software pipelines. His background spans materials science, symbolic regression, equivariant ML and LLMs grounded in real physical domains, with a standing commitment to declarative, deterministic infrastructure across every deployment.
Daria owns Alacene's retrieval and trust layer: RAG architectures that know when to abstain, uncertainty scoring, and evaluation frameworks for factual QA. Her published work on adaptive retrieval, question classification and RAG evaluation is the backbone of how Alacene keeps answers grounded in the source.
Maxim works across modelling and infrastructure at Alacene, designing the on-premise retrieval stack. Twenty years across probability theory, Bayesian methods and deep learning, with deep expertise in reproducible, declarative infrastructure (NixOS, systemd, microVMs) for environments where AI has to run inside the customer's perimeter — not behind someone else's API.