Probably the best chatbot framework in the world.
Deployed amongst others on Work in Romania and Florida International University
That is the reaction to the first demonstration. It is also the smallest thing the framework does.
Prodigy - Applied AI Research & Engineering
One framework beneath every system.
Prodigy is an applied AI lab combining research, architecture, product thinking and software engineering. At its core is AiRox, our proprietary AI framework for building and operating intelligent systems across government, critical missions and enterprises.
One framework · AiRox
2018Founded · Cluj-Napoca
WorldwideMarkets
01 · The core framework
AiRox - the proprietary foundation powering Prodigy’s AI systems.
AiRox brings together the technologies required to build, operate and evolve intelligent systems, from models and agents to knowledge, multimodal interaction, workflows, infrastructure and human oversight.
00
AiRox LLM + proprietary models
At the core of AiRox is our proprietary language model, designed for controlled, adaptable and private AI deployments. It can be specialised through fine-tuning and domain knowledge, deployed on-premise when required, and orchestrated alongside other leading models within the broader AiRox ecosystem.
01
Third-party models
Different tasks require different intelligence. AiRox can work across language, reasoning, vision, voice and specialised models, selecting the right capabilities for each system and use case.
02
Agents + orchestration
Models become coordinated systems. AiRox manages agents, roles, routing, evaluation, tool use and interactions between specialised components rather than relying on a single model for every task.
03
Knowledge + RAG
Organisational knowledge becomes accessible to AI through approved sources, documents, retrieval, semantic search and structured knowledge bases.
04
Context + memory
Intelligent systems need continuity. AiRox maintains contextual and persistent information across interactions, users and workflows, so systems can operate with more than the information contained in a single prompt.
05
Voice + vision
AiRox extends beyond text with speech, image and video capabilities - including speech-to-text, text-to-speech, computer vision and multimodal intelligence.
06
Data + tools
AiRox connects intelligence to the systems around it. APIs, operational data sources, databases and external tools become available to models and agents as part of the same architecture.
07
Workflows + automation
AI output can become operational action. AiRox supports workflows, approvals, automations and integrations that connect intelligence directly to organisational processes.
08
Governance + human control
Access, traceability, evaluation, cost visibility and human validation are built into the system. AI can generate context, recommendations and actions while defined controls remain with the organisation and its users.
09
Deployment + AI infrastructure
AiRox can operate across public, private, hybrid and on-premise environments. Its architecture can be designed around the models, compute, GPU infrastructure, networking, storage, latency and data-control requirements of each deployment.
Public cloud
Private cloud
On-premise
Hybrid
One framework. Many missions.
The full framework02 · The technology chain
Prodigy covers three essential levels of the AI economy.
The application defines the task. The task defines the model. The model defines the compute.
01
Energy
The energy capacity that powers digital infrastructure.
02
AI compute infrastructure
GPUs, servers, storage, networking and data centres.
Prodigy experience03
Foundation models
Language, reasoning, voice, vision and specialised models.
Prodigy experience04
AI applications and systems
Products that connect models to real data, processes and objectives.
Prodigy experienceWe connect models, infrastructure and applications into complete AI systems built for real-world operation.
03 · Positioning & how we work
An applied AI lab for complex, real-world systems.
Prodigy combines proprietary AI technology with research, architecture, product and software engineering to understand complex problems and turn them into operational systems.
We start with the problem, define the right combination of models, data, infrastructure and workflows, and carry that architecture through to deployment and operation.
This is what allows the same AiRox foundation to be adapted to very different domains, environments and mission requirements.
/ 01
Research
Understand the problem before committing to a build.
/ 02
Architecture
Models, data, infrastructure and control.
/ 03
Product
Workflows and real use, not demonstrations.
/ 04
Engineering
Built, integrated, operated.
How we work
From ambiguous problem to operational capability.
Six stages. Each one produces evidence, and evidence is what makes the next stage safe to fund.
1Frame
The mission, the users, the constraints and the decision that has to improve.
Problem definition
2Research
Workflows, data, models, integration, security, feasibility.
Research brief
3Prototype
AiRox against the riskiest assumption. Working software, not slideware.
Validated direction
4Productise
Experience, workflows, governance and deployment architecture.
Mission-ready system
5Deploy
Real systems, real data. Cloud, private cloud or on-premise.
Operational capability
6Evolve
Measure the use, improve the models, find what becomes a product.
Growing product IP
The starting point is never a model. It is the mission, the people who carry it out and the decision that has to improve as a result.
04 · Portfolio
One technology foundation. Many missions.
Prodigy’s portfolio spans government, critical operations, enterprise and public-facing systems across multiple industries and markets.
Each project addresses different problems, but they share the same foundation: AiRox, continuously extended through real-world use.
Deployed
Government AI · public administration
Work in Romania
Official immigration knowledge, answered in the languages the people asking it actually speak.
In development · confidential
Public safety
ARGUS
AI-supported investigative intelligence combining relationships, geospatial data, movement patterns, scenario analysis and prediction, turning fragmented information into structured operational context. Analysis remains human-led throughout.
In development
Computer vision · real-time intelligence
VIGIL
Real-time image and video intelligence for detecting events, correlating observations and transforming visual streams into operational information that can support human decision-making.
In development
Mission-critical AI · multi-sensor intelligence
Midbrain / AERO
An operational intelligence system combining cameras, sensors, data and AI to build situational awareness, identify relevant events and support human decision-making in complex environments.
Delivered
Knowledge · health · learning
NIOSH
A family of AI-supported institutional systems spanning occupational health and safety, digital health, knowledge management and multimodal learning experiences.
In development
Defence & operations
NIAG Command Center
A command-and-control environment bringing live video, mapping and operational tasking together across drones, robotic platforms and field teams, creating a shared operational picture from distributed assets and data.
Delivered
Government · India
Government of Maharashtra
A large-scale human-resources platform delivered for a state administration in India, operating across the public-sector employee lifecycle.
Multilingual by design, multimodal by default. Where a system answers questions, it does so from an approved body of knowledge and can cite the source behind every response, switching language within the same session. Users move between text and voice, and context is preserved across languages, channels and interactions. These are not generic chatbots wrapped in institutional branding; they are controlled, knowledge-grounded systems built to operate reliably in real organisations.
WCO·
NATO / NIAG·
OECD·
NVIDIA
Institutional and ecosystem relationships. The NVIDIA relationship is an ecosystem relationship within accelerated computing for AI.
More assets across public administration, critical missions, enterprise and knowledge, each labelled with its real maturity.
The full portfolio05 · Why Prodigy
Research-lab discipline. Product-company speed.
Mission-critical responsibility.
Research-led
Assumptions are tested in research before a build is committed to, because correcting a wrong assumption at that stage costs a fraction of what it costs later.
Product-minded
Systems are shaped around adoption and measurable value, and are built to be operated rather than demonstrated.
Framework-powered
AiRox carries proven architecture forward from one project to the next, so each system begins from a larger base of existing capability.
Mission-ready
Security, traceability and human oversight are designed in from the start, because they cannot be added convincingly to a system that was not built for them.
Contact
Start with the problem, not the technology.
Send us the challenge, the constraints you are working within and the outcome that has to improve. We will come back with what a working system would involve, how it would be built and operated, and whether AI is the right tool for the problem at all.
Start the conversationHeadquartersCluj-Napoca, Romania