AI & Machine Learning Company in Halifax

Two things set a Halifax ML number: the state of your data, and how far past a demo you need to get. Proposals are fixed-fee rather than published bands or open T and M estimates. A scoped Halifax AI proof of concept is a six to ten week engagement covering data audit, a baseline model, and a hosted demo on AWS ca-central-1 or Azure Canada Central. A custom production ML model (hydroacoustic anomaly classifier, AIS vessel detection, lapse propensity scorer, claims triage model, fisheries forecaster) adds MLOps, monitoring, and a model card aligned to OSFI Guideline E-23 or Canada's Directive on Automated Decision-Making. Full production AI systems add multiple models, defence export control screens, and Canada's Ocean Supercluster reporting templates. Halifax sits below Toronto and Montreal and reflects the Dalhousie, Saint Mary's, and NSCC talent baseline plus the lower Atlantic Canada cost-of-living adjustment. We give fixed-fee proposals rather than open T and M estimates.

OSFI Guideline E-23 on Enterprise-Wide Model Risk Management, published by the Office of the Superintendent of Financial Institutions, sets the model risk governance standard for federally regulated financial institutions including the Big Six banks, life and P&C insurers like Manulife, and federal credit unions. The 2024 revision broadened scope to include AI and machine learning models and reinforced expectations around model inventory, development standards, validation independence, ongoing monitoring, and challenger model testing. For Halifax insurance back-office builds we map every AI model to the E-23 model lifecycle: design and development with documented assumptions, independent validation by a second team or the client's model risk function, ongoing monitoring with drift detection and performance triggers, and a challenger model testing programme. Manulife, John Hancock, Empire Life, and the Atlantic credit unions all evaluate against this template, and our model card output drops directly into the client's model inventory.

Canada's Ocean Supercluster, headquartered downtown on Hollis Street, has received more than $300M in federal funding since 2018 to accelerate Atlantic ocean tech commercialisation, and runs programmes like the Ocean Startup Project, the Innovation Ecosystem programme, and project-based co-funding for member-led initiatives. For OSC-funded engagements we structure the workplan to match the OSC project agreement (milestones, deliverables, IP terms, member contribution tracking), deliver the technical narrative and time-tracked logs in the OSC reporting template, and coordinate with the member organisation's finance and grants team on claims and reimbursement. We have worked or studied patterns on OSC member projects across hydroacoustic processing, AUV autonomy, fisheries monitoring, and offshore wind site assessment. The Ocean Frontier Institute at Dalhousie, COVE in Dartmouth, and BlueNode often appear as co-applicants or technical partners on the same projects.

Yes, under the Controlled Goods Programme, the Defence Production Act, and the Export and Import Permits Act. The Royal Canadian Navy operates from CFB Halifax with Maritime Forces Atlantic at HMC Dockyard, and Irving Shipbuilding's Halifax Shipyard delivers the Canadian Surface Combatants (Type 26-derived) and the Harry DeWolf-class Arctic and Offshore Patrol Ships under Canada's National Shipbuilding Strategy. Sonobuoy and ASW supply runs through Ultra Maritime in Dartmouth. For defence-touching AI work we register our team and facilities under the Controlled Goods Programme where required, screen every engagement for ITAR-controlled technical data, structure the supply chain and team to keep controlled data inside cleared environments, and coordinate with the client's security officer on the Project Security Instructions. We do not accept work we cannot legally execute, and we map every export-controlled component before week one wraps.

Atlantic Standard Time runs one hour ahead of Eastern Standard Time during standard time and aligns with Eastern Daylight Time during daylight saving. For Halifax clients that work with Toronto, Montreal, and New York counterparts, this means Halifax teams pick up calls before the rest of the EST corridor starts the day, which suits Maritime back-office operators like Manulife and John Hancock that run early actuarial shifts and ocean tech operators that align with European customers in CET. For our delivery model the AST overlap with Chandigarh (typically 8.5 to 9.5 hours ahead during standard time) and Edmonton (typically 4 hours behind during standard time, 3 hours during daylight) gives a sustained morning Halifax window for synchronous standups, design reviews, and OSFI Guideline E-23 model validation walkthroughs. Halifax clients get a real product partner in their morning, not an overnight ticket queue.

We default to AWS ca-central-1 in Montreal (roughly 1250 kilometres west, three Availability Zones), Azure Canada Central in Toronto and Azure Canada East in Quebec City for backup and DR, and Google Cloud northamerica-northeast1 in Montreal and northamerica-northeast2 in Toronto for ML and data warehousing. For LLM layers that must stay in Canada (PHIA Nova Scotia health data, OSFI-regulated banking and insurance, federal Protected B workloads) we use Cohere's Canadian endpoints, Anthropic Claude through AWS Bedrock in ca-central-1, GPT-4o through Azure OpenAI Service in Canada East and Canada Central, or self-host Llama 3.3 and Mistral on Canadian GPU instances. Cross-border processing is acceptable only when a Privacy Impact Assessment signs off and the client's privacy officer documents the basis under PIPEDA section 4.1.3 and the Nova Scotia Personal Information International Disclosure Protection Act. We document residency in the model card and the data processing agreement so PIPEDA, FOIPOP, and PHIA auditors have a clear answer.

A typical Halifax insurance back-office model (lapse propensity, claims triage, NIGO document detection, agent productivity scoring, fraud screening) takes fourteen to twenty-two weeks from kickoff to production, assuming clean historical data and OSFI Guideline E-23 documentation as part of scope. Week 1 to 4 is data audit and baseline. Week 5 to 12 is modelling, iteration, and challenger testing. Week 13 to 18 is MLOps, monitoring, shadow mode deployment, and internal audit review with the client's model risk function. Week 19 onward is gradual rollout under the E-23 ongoing monitoring programme. If you are pre-data or need labelling, add six to eight weeks. We have shipped to this cadence inside Canadian OSFI-regulated stacks for life insurance, P&C, and credit union clients, and Manulife, John Hancock, and Empire Life all evaluate against the same template.

Most custom AI work qualifies for the Scientific Research and Experimental Development (SR&ED) program, which returns 35 percent refundable credit on eligible R&D spend for Canadian-controlled private corporations (CCPCs) up to the $3M expenditure limit, and 15 percent non-refundable above that. Eligible activity generally includes novel modelling, architecture experimentation, algorithmic uncertainty, and systematic investigation, not routine integration. Atlantic Canada Opportunities Agency (ACOA) programmes including the Regional Economic Growth through Innovation (REGI) stream, the Atlantic Innovation Fund successor programmes, and Business Development Programme grants frequently stack with SR&ED for Halifax SMEs. Canada's Ocean Supercluster co-funding stacks for member-led ocean tech projects. We deliver time-tracked logs, technical narratives, experiment histories, and failed-hypothesis documentation aligned with the CRA T661 form and the relevant ACOA or OSC reporting template. Final eligibility sits with your SR&ED consultant and the CRA.