Technology Stack
The platforms we build on.
Darius and the AIDARIUS team do not recommend AI platforms they have not used in production. Every tool below is one we have deployed with real clients — and understand well enough to tell you when not to use it.
Core platforms
Five ecosystems. Deep expertise.
These are the platforms we train clients on, integrate into business workflows, and draw on when designing AI strategy. Each one is evaluated independently — the goal is always fit, not loyalty to a vendor.
Enterprise AI
Microsoft
Microsoft is the dominant AI platform for organisations already running on Microsoft 365. We deliver in-depth Copilot training, Copilot Studio deployments, and Azure AI integrations — helping teams move from basic prompting to genuine productivity transformation. With 25+ years in enterprise IT, Microsoft's ecosystem is the one we know deepest.
Microsoft has pivoted to an agent-first platform — computer-using agents are now live in Copilot Studio, and Azure AI hosts a multi-model catalogue spanning open and proprietary models. The Phi family of open-weight models (MIT licence) extends Microsoft's reach into on-premise and edge deployments. Our engagements span readiness assessments, end-user training, EU AI Act governance frameworks, and custom Copilot agent builds.
AI Assistant & API
Anthropic — Claude
Claude is our primary AI assistant for strategic work, content development, and complex reasoning tasks. The Claude Opus family consistently ranks at the top of independent AI intelligence benchmarks — the most capable models available for multi-step reasoning, analysis, and knowledge work. The Claude Sonnet family handles production workloads at speed. As certified Claude power users and trainers, we help organisations deploy Claude for knowledge work, document processing, and internal AI agent builds. Claude's safety profile and constitutional design make it the first-choice model for regulated industries and governance-conscious clients.
Generative AI
OpenAI
OpenAI's GPT family represents the current generative AI standard, with full-scale flagship models for complex reasoning and compact mini / nano variants bringing frontier performance to cost-sensitive workloads. We use OpenAI across training programmes, custom GPT builds, and API-based automations — and teach teams how to evaluate outputs critically rather than accept them uncritically.
AI Ecosystem
Google — Gemini
For organisations running Google Workspace, Gemini represents the most natural AI entry point. We train teams on Gemini for Workspace — Docs, Gmail, Meet, Sheets — and advise on Vertex AI deployments for more advanced use cases. The Gemini Pro family delivers class-leading multimodal and long-context performance, making it particularly relevant for content-heavy and research-driven organisations.
AI Infrastructure
Nvidia
NVIDIA underpins the compute layer of modern AI — from cloud inference to on-premise deployments. The Nemotron family delivers efficient open models purpose-built for agentic AI applications. NIM microservices enable containerised deployment of leading open-weight models on NVIDIA-accelerated infrastructure. We advise clients on infrastructure decisions and how to evaluate GPU-backed platforms for privacy-sensitive or high-volume workloads.
Enterprise AI Platform
IBM — watsonx
IBM watsonx is the enterprise AI and data platform built for regulated industries — banking, insurance, healthcare, and government — where explainability, auditability, and data governance are non-negotiable. The IBM Granite model family (open-weight, Apache 2.0) provides transparent, auditable models across a range of sizes — well-suited for organisations that require full visibility into model behaviour. We advise organisations on IBM's AI governance tooling and Granite deployments for clients with EU AI Act compliance requirements or on-premise data mandates.
Open source
Open-source models & frameworks.
Proprietary APIs are not the only answer. For privacy-sensitive workloads, local deployments, or clients who want full data sovereignty, open-source models and frameworks are first-class options.
Open-Source LLM
Meta Llama
The Llama 4 family marks a generational leap in open-weight AI — bringing native multimodal capabilities, Mixture-of-Experts architecture, and massive context windows to models that can run fully on-premise. For organisations that cannot send data to the cloud, the latest Llama family is now a credible alternative to proprietary APIs on most benchmarks.
European Open LLM
Mistral AI
Mistral's latest Small family consolidates reasoning, multimodal vision, and agentic coding into a single unified model. The Mistral Large family handles larger-scale workloads. Both are available with European data residency, making Mistral the natural choice for EU organisations with GDPR and EU AI Act compliance requirements. We recommend Mistral as the default European alternative to US-hosted frontier models.
Local Model Runner
Ollama
Ollama makes running open-weight models locally as simple as a single terminal command. For workshops, demos, and proof-of-concept environments, we use Ollama to show clients exactly what on-device AI looks like — no API keys, no cloud costs, no data leaving the machine. A practical introduction to local AI for technical and non-technical audiences alike.
Model Hub & Ecosystem
Hugging Face
Hugging Face is the de facto hub for open-source AI models, datasets, and tools. We use it for model evaluation, fine-tuning workflows, and Spaces demos. For clients evaluating which open-source model fits a specific task — summarisation, classification, embeddings — Hugging Face is the reference point for benchmarks and reproducible comparisons.
Agent Framework
LangChain & LangGraph
LangChain is the most widely used framework for building LLM-powered applications and agents. LangGraph extends it to stateful, multi-step agent workflows with human-in-the-loop control. We use both when designing AI agent architectures for clients — from simple RAG pipelines to complex multi-agent orchestration.
For organisations moving beyond basic prompting into AI-powered automation — document processing pipelines, research assistants, internal knowledge bots — LangChain and LangGraph provide the scaffolding. Combined with open-source models, they enable fully private, fully controlled AI systems with no vendor lock-in.
Our approach
Vendor-neutral. Results-first.
AIDARIUS has no obligation to any AI vendor. Every recommendation starts with your organisation's actual needs, infrastructure, and team capacity — not a preferred product.
Platform-fit assessment
Before recommending any platform, we evaluate your existing tech stack, licensing, data residency requirements, and team skill level. The best AI tool is the one your team will actually use.
Governance from day one
Every deployment includes usage policy, data handling guidelines, and EU AI Act compliance considerations — not as an afterthought, but as part of the initial design.
Measured outcomes
Adoption metrics, time-saved benchmarks, and skill assessments are built into every programme. If we cannot measure it, we have not solved it.
Work with us
Not sure which platform fits your team?
A 90-minute AI Strategy Session maps your organisation's needs to the right tools — with no vendor agenda.