Software that reasons, acts and explains itself.
inMotionMedia is a software and applied-AI company. We design agentic systems that model their domain, plan the evidence they need, use tools safely and leave an auditable trace behind every decision.
From machine learning to systems that can be held accountable.
We combine applied AI, knowledge representation and classic software engineering — so that intelligent behaviour is also predictable, inspectable and safe to run.
Applied AI & machine learning
Intelligent systems in production: extraction, classification, computer vision, embeddings and hybrid semantic search.
LLM & agentic systems
Tool-using agents and multi-agent orchestration on a deterministic runtime, with supervision and recovery of long-running work.
Knowledge representation
Knowledge graphs and entity–relationship models that keep claims, evidence and context explicit and queryable.
Decision support & reasoning
Rule engines, probabilistic updating, planning and replanning — systems that state their confidence and abstain when evidence is not enough.
Data engineering
Heterogeneous sources — APIs, bank feeds, invoices, mailboxes, documents, the web — turned into clean, isolated pipelines.
Trustworthy & explainable AI
Human approval gates, auditable decision traces, benchmarking, EU data residency and GDPR by design.
Research & development
An in-house R&D team that takes ideas from experiment to prototype to product — and runs every system on our own company first, before it reaches anyone else.
OMNI — our agentic operating platform.
Built in-house and used every day to run our own company: finance, infrastructure, hosting, marketing and software delivery. OMNI is where our research turns into production.
- 0150+ deterministic tool modules, each with a declared capability contract.
- 02Every agent action passes a gatekeeper — identity, permissions and audit.
- 03Irreversible or outbound actions require explicit human approval.
- 04Profile-scoped isolation: one codebase, many fully separated deployments.
- 05A complete decision and usage trace, including the cost of every model call.
- 06EU-hosted models wherever data leaves the house.
The loop we build into every system: claims, evidence and context are modelled explicitly; the agent plans, acts through tools, updates its belief and either decides — with a trace — or asks for more.
Agents that verify before they conclude.
Our R&D focuses on the hardest part of agentic AI: knowing what is true, how sure we are, and why.
R1Cognitive architectures & world models
State representation, memory and reasoning interfaces that model a verification task as claims, evidence and context — a world model the agent can reason over, not a flat prompt.
R2Evidence graphs
Graph representations linking every claim to its supporting and contradicting evidence, its source and its context, so conclusions can be inspected and challenged.
R3Adaptive evidence planning
Agents that decide which evidence to gather next, orchestrate the right tools and replan dynamically when a source fails or the picture changes.
R4Uncertainty, abstention & traces
Probabilistic, Bayesian belief updating, calibrated confidence, principled abstention — and auditable decision traces that an auditor can read and an engineer can replay.
European collaboration. We join European research and innovation consortia as a technology partner for cognitive architectures, evidence reasoning and trustworthy agentic AI — from research prototype to working software.
Systems we build and run.
OMNI
Agentic operations platform: multi-agent task execution, 50+ tool modules, human-in-the-loop approvals and a full audit trail. Runs our company every day.
Addie 2.0
From a single URL to market research, buyer personas and on-brand creative — an agentic marketing pipeline with human review at every stage.
Finance automation
Bank APIs, invoicing and deterministic payment matching that reconciles receivables across sources without manual work.
Deterministic where possible. Probabilistic where necessary. Human where it matters.
Model
Represent the domain — entities, relations, claims, evidence and context.
Plan
Decide what evidence is needed and which tools can obtain it.
Act
Execute through audited tools, update beliefs, replan when reality differs.
Account
Decide with stated confidence — or abstain — and leave a trace behind.
Let's build something that can explain itself. info@inmotionmedia.sk
Záhradnícka 7
811 07 Bratislava, Slovakia
Commercial Register, Municipal Court Bratislava III, Sro 56173/B
Founded 2009