Research papers
Submitted for publication where the work warrants it. Technical reports for the rest.
Engage · Research & Development
Turning research into competitive advantage.
Active research in emerging technologies — funded under the Australian Government R&D Tax Incentive.
01 What this is
01 What this is
Datakey is a research-first firm with active engagements spanning distributed systems, graph neural networks, software and neural architectures, on-device AI, zero-boundary databases and GPU computing.
If you have an idea or project that needs structured exploration, experimentation, and new knowledge to validate, a partnership with us can deliver your project at substantially reduced costs.
02 The R&D Tax Incentive
What does co-funded research look like?
The Australian R&D Tax Incentive is a refundable tax offset for qualifying research and development. For eligible companies the scheme returns up to 43.5% of qualifying R&D costs, substantially reducing your project bottom line.
The Rule
43.5%
refundable tax offset on qualifying R&D for eligible Australian companies.
Working Example
Total project$400,000
Qualifying R&D portion75%
Eligible spend$300,000
Refundable offset (43.5%)~$130,000
Project bottom line~$270,000
Indicative only.
Datakey has 3+ years of structured research under this scheme applied to internal and client engagements. Done well, this approach makes ambitious research affordable and deepens the technology itself.
In practice
03 How a research engagement works
The primary pathway: the client funds the R&D project, structured to qualify for the scheme. Datakey contributes research depth and technical execution.
Engagement scope ranges from focused research efforts over 1–3 months through to multi-year programmes with deepening outcomes.
Public
contributions
contributions
Our preferred engagement. Where research is publicly valuable, we structure the research as a public contribution via peer-reviewed publications and open-source releases. Same engagement structure, different community outcome.
What you receive
Open-source
Repositories, libraries, and reference implementations as open source contributions.
IP on prototypes / MVPs
The technical deliverables the research facilitated.
R&D Tax Incentives
Registrations and the technical record for your tax claim.
04 Where our research sits
Our active research topics.
Software architectures of intelligent systems
The engineering layer beneath distributed intelligent systems. Systems-level architecture for coordination at scale.
Multi-model neural architectures
Multiple specialised models coordinating through joint embeddings or common datasets for shared focus.
Distributed intelligent systems
Intelligence coordinating across nodes, devices, and organisational boundaries — browser, device, edge, and cloud — in real-time.
CPU-GPU unified memory architectures
Apple Silicon unified memory, zero-copy CPU/GPU compute, on-device inference at scale.
On-device and edge AI
On-device intelligence where privacy, locality, and autonomy are primary constraints. Intelligence pipelines running on user hardware.
Zero-copy and zero-boundary databases
Database engines with zero serialisation from wire to both CPU and GPU computing bounds.
05 Recent research
- Data Structures for GPU and CPU in Apple Silicon Unified Memory Whitepaper
Foundational research toward a new database architecture, where CPU and GPU operate on the same physical bytes for massively parallel GPU-level query execution. Read the whitepaper →
- Zero-Boundary Database Architecture In progress
Benchmarks paper measuring throughput, atomic operations, and query performance against incumbents — Redis, Kafka, Postgres, DuckDB, pgvector, Chroma, Qdrant, Milvus. Project · mgraph →
- Distributed Coordination for Real-Time Intelligent Systems FY24/25
Zero-copy methodologies, decentralised permission structures, and cryptographic verifiable lineage for high-throughput graph synchronisation.
- Graph Neural Networks for Multi-Agent Coordination FY23/24
Cooperative frameworks where multiple specialised models coordinate through graph-based representations, applied to AI productivity SaaS at scale. trudi.ai →
- Knowledge Architectures for Intelligent Retrieval FY23/24
Graph-based knowledge representation, retrieval-augmented generation over structured ontologies, and domain-aware search applied to AI productivity SaaS. trudi.ai →
Works with
Other engagement models
01
Strategic Advice
Research depth surfaced as judgement on strategic direction before technical work begins.
Read more →02Engineering & Development
Research output implemented as paid delivery on top of the substrate the work produced.
Read more →03Partnerships & Equity
Same engagement structure with closed-IP outcomes and shared upside pathways.
Read more →Capabilities
01
AI, ML & Neural Architectures
Where most active research lives: graph neural networks, on-device inference, distributed coordination.
Read more →02Architecture & Infrastructure
Where systems-architecture thesis becomes delivery capability: wire formats, transports, runtime layout.
Read more →03Technology Strategy
Why we research what we research, in this order, and how IP is structured around it.
Read more →Contact
Tell us what you're working on.
We reply within two business days. If a call would be faster, book a thirty minute conversation.
Prefer to talk?
Book a thirty-minute call