The Consortium
xSeraAI is pursuing CRC-P Round 19 funding to build Australia's sovereign domain-specific AI capability. The consortium is built on pre-existing research and industry relationships — not cold introductions assembled for a grant application.
The Partners
Every entity in this consortium has an established working relationship with xSeraAI or the founding team. The trust required for deep industry data sharing already exists.
ATHENA CORE architecture, DSLM training methodology, and IP holder. Nathan Nguyen Luu (Founder & CEO) leads the consortium as the primary applicant. xSeraAI owns the data curation pipeline, model evaluation framework, and commercialisation pathway for sovereign AI outcomes in Australian industry.
Motor Trades Association of Queensland. 6,000+ automotive businesses across Queensland. 11 specialist industry divisions covering passenger vehicles, trucks, motorcycles, and marine. Eight-year relationship with the xSeraAI founding team provides the trust required for deep, proprietary data sharing. Primary data source and 200-business pilot deployment partner for the automotive DSLM.
Prof. Michael Milford — ARC Laureate Fellow, QUT Centre for Robotics. Nine-year relationship with the xSeraAI team. Prior CRC project experience. Leads DSLM architecture validation, federated learning privacy framework development, and model evaluation methodology across Work Packages 1, 2, and 3. QUT's applied AI expertise grounds the research in peer-reviewed, defensible methodology.
Dr. Sue Keay. H2O AI 100. Former Chair, Robotics Australia Group. National AI policy alignment and sovereign AI strategic positioning. Dr. Keay's cross-sector perspective connects xSeraAI's narrow-deep methodology to Australia's broader AI policy agenda — ensuring the CRC-P application aligns with National AI Plan priorities and sovereign capability objectives.
Luke Kelleher (Founder). 175+ live AI agents deployed across Australian SMEs — handling calls, bookings, and CRM for automotive workshops, professional services, and trade businesses. SophiieAI is not simply the delivery channel for xSeraAI models. It is the data engine. Every Sophiie interaction generates structured, industry-classified data that feeds directly into DSLM training pipelines. Without this deployment network, the narrow-deep thesis has no proprietary data pipeline. Without xSeraAI's DSLM methodology, Sophiie's data remains unstructured and underutilised.
Dr. Brett Dale (President, AMA Queensland). The healthcare vertical bridge connects the CRC-P Round 19 automotive proof-of-concept to the CRC Round 28 healthcare expansion. AMA Queensland's involvement signals clinical credibility, provides access to Medicare, TGA, and AHPRA compliance frameworks, and positions xSeraAI for QLD Health alignment as the methodology scales beyond automotive.
Strategic Partnership · SophiieAI
The SophiieAI partnership is the most strategically significant element of the xSeraAI consortium. It transforms the DSLM thesis from a research concept into a continuously improving system grounded in real Australian industry data.
Customer-facing AI agents deployed 24/7 across automotive, trades, and professional services — generating real-world interaction data at scale.
ATHENA CORE processes each interaction: intent classification, domain categorisation, resolution path, compliance flags. Structured signal — not raw transcripts.
The structured signal feeds directly into DSLM training pipelines — real Australian industry data that no foundation model has access to, and no competitor can replicate.
Improved DSLMs make Sophiie agents demonstrably more accurate on domain-specific queries. Higher accuracy drives retention. More deployments produce more training data. The loop compounds.
This partnership isn't xSeraAI using Sophiie. It's a co-dependency where both get exponentially more valuable together. Sophiie generates the data. xSeraAI structures it into sovereign AI capability. The flywheel only works if both sides are operating.
Two-Vertical Strategy
The two-vertical structure is deliberate. CRC-P Round 19 funds the automotive proof-of-concept. Success there validates the DSLM methodology for the CRC Round 28 application — a $50M healthcare AI expansion with national implications.
Australia's first narrow-deep DSLM built for the automotive industry. Trained on the proprietary data that defines the sector — from AUR competency frameworks to OEM fault codes to live workshop operational data from Sophiie deployments.
The methodology proven in automotive applies directly to healthcare — a domain equally dependent on proprietary, compliance-heavy, sovereign data. AMA Queensland bridges the two verticals through Dr. Brett Dale's involvement from CRC-P Round 19.
Research Agenda
Five work packages structured across a 24-month research timeline. Each package produces replicable, peer-reviewed methodology — not proprietary black-box outputs. The research outputs belong to the consortium and feed directly into commercial deployment.
Developing the end-to-end pipeline for ingesting, cleaning, classifying, and structuring proprietary industry data for DSLM training. Includes domain ontology construction, data governance framework, and privacy-preserving curation protocols.
Architecture selection, fine-tuning methodology, and evaluation benchmarking for domain-specific language models. Produces peer-reviewed DSLM evaluation framework applicable across verticals — automotive first, healthcare next.
Federated learning architecture enabling model training across distributed industry data without centralising sensitive records. Includes differential privacy implementation and compliance assessment against Australian Privacy Act requirements.
Real-world deployment of the automotive DSLM across 200 MTAQ member businesses via SophiieAI's agent platform. Measures accuracy gains, workflow integration, compliance performance, and training signal quality from production interaction data.
IP structuring, licensing framework, commercialisation pathway, and CRC Round 28 application preparation. Ensures research outputs translate into sovereign Australian AI capability — not offshore-owned technology applied to Australian data.
Market Validation
Domain-specific language models are not a research hypothesis — they are the direction the global AI market is already moving. xSeraAI's thesis is validated by market evidence; the Australian sovereign layer is what remains to be built.
Legal DSLM valued at $11B — the clearest global proof that narrow-deep, domain-specific AI commands premium valuation over general-purpose models. Harvey does for law what xSeraAI does for automotive and healthcare.
Gartner projects that more than 60% of enterprise GenAI deployments will use domain-specific models by 2028. The market is moving from general-purpose toward industry-native. The question is who builds the Australian sovereign layer.
DSLMs consistently outperform GPT-4o class models by 25–30% on clinical and specialist industry tasks. The performance gap is not marginal — it is the difference between a tool practitioners trust and one they cannot deploy in compliance-sensitive contexts.
Get Involved
CRC-P Round 19 closes 12 May 2026. If your organisation has domain-specific data, research capability, or industry reach in automotive or healthcare, contact us before the deadline.