DimechainQuantum neural networks
Quantum and hybrid model development using frameworks including PennyLane, Qiskit Machine Learning and sQULearn.
Dimechain × Automatski
AI research & enterprise applications
Applied AI delivery at Dimechain, supported by Automatski’s research into efficient and alternative computing architectures.
01 / Post-transformer research
A smaller computational footprint.
Automatski’s Post Transformer AI Venture investigates how quantum layers and alternative architectures could reduce the compute and memory required by AI. Its brief describes experiments based on Llama 2, 3 and 4, replacing selected layers while keeping the surrounding architecture.
The goal is more accessible training and inference. An evaluation must compare model quality, memory, latency and total compute on the same task and hardware. This research does not establish independently reproduced performance, a production service or customer pricing.
Discuss an AI research pilot ↗02 / Partner research portfolioResearch program Research program Research implementation Reference implementation Probabilistic computing Neuromorphic research Cognitive computing Developer tool
Several paths to intelligence.
Quantum machine learning
Machine-learning workflows designed to connect familiar development interfaces with quantum execution backends.
Quantum Transformers
Models designed to represent quantum-mechanical problems in physics and chemistry.
Quantum Large Language Model
A training and inference pipeline described for a tensor-network simulator or Automatski’s quantum backend.
Pithanos
A probabilistic-programming research framework for uncertainty-aware modeling and inference.
Akida (Automatski)
Spiking neural networks for event-based perception, monitoring, robotics and edge computing. The research also explores reinforcement-learning integrations.
Diastaseis
Hyperdimensional computing and vector symbolic architectures for reasoning, classification, memory, graph processing and hybrid AI.
Curiosity
A terminal-based AI coding agent designed to run locally on a laptop.
Discuss access ↗03 / Dimechain delivery
Built around your workflow.
Dimechain develops B2B and on-demand AI solutions around an organization’s data, users and operational constraints. Engagements can include retrieval and assistants, process automation, decision support and product integration. Research-led components enter a project through a scoped evaluation.
Define the workflowEstablish the baselineEvaluate a pilotIntegrate and operate
Healthcare AI / Research conceptsResearch concept Research concept
Connect data. Support research.
Our earlier quantum portfolio introduces DIM-12 and RadiologyGPT as healthcare AI research directions. The proposed work combines medical imaging, pathology and genomic information with deep learning, quantum neural networks and reasoning methods.
DIM-12
A proposed oncology AI framework exploring multimodal data, cancer-related pattern analysis, explainability and uncertainty. Architecture descriptions include a layered deep-learning model with quantum and neuro-symbolic components.
RadiologyGPT
A proposed multimodal assistant for imaging analysis and report preparation, with research into combining images, text and audio and integrating with clinical information systems.
These are research descriptions. Clinical validation, regulatory clearance and diagnostic performance have not been established by the material reviewed here. They are not presented for clinical diagnosis.