/01 ETL · QUALITY · LINEAGE

Data Engineering & Governance

Structure, harden and make your data usable. Pipelines, modelling, quality checks and lineage — so your data becomes a reliable, shareable asset, integrating with your existing stack rather than replacing it.

Who it’s for

Teams facing scattered data, quality or reliability issues, or a migration to a more modern platform. Leaders who want to bring order to their data flows.

Methodology

A diagnostic of what exists, iterative design, integration with your current stack — no rip-and-replace, no big-bang. Deliverables are tested and documented as we go.

Deliverables

  • Source and flow mapping (data lineage)
  • Robust, documented data pipelines (ETL/ELT)
  • Data modelling and structuring
  • Automated quality checks (tests, alerts)
  • Migration support, with no data loss
  • Tools made accessible to non-technical users

Typical use cases

  • Harden and unify scattered data assets
  • Set up quality checks and data lineage
  • Build pipelines that feed BI and dashboards
  • Support a data migration while preserving integrity

Stack & tools

  • Python
  • SQL
  • dbt
  • Apache Airflow
  • Kestra
  • Databricks
  • BigQuery
  • PostgreSQL
  • Great Expectations
  • Collibra
  • Dataiku
  • Docker
  • Kubernetes
  • AWS
  • Azure
/02 RAG · AGENTS · MCP

Applied Artificial Intelligence

Putting AI to work, concretely. Design and deployment of bespoke RAG solutions, agents and integrations — using, orchestrating and optimising existing models, self-hostable if you wish. The goal: genuinely useful tools, not demos that sleep in a notebook.

Who it’s for

Organisations that want to use generative AI on their own data and processes: an internal copilot, augmented search, agents and automation. Teams that want reliable uses rather than hype.

Methodology

A fast prototype on a concrete case, iterations with you, then a pragmatic rollout. We favour self-hostable models so you keep control of your data.

Deliverables

  • Copilots and augmented search (RAG) on your documents
  • AI agents and automation connected to your tools
  • MCP integrations and tool-use for models
  • Prompt, context and “harness” engineering
  • Deployment of self-hosted (open-weight) models
  • Use-case scoping and good practices

Typical use cases

  • An internal copilot that answers from your documentation
  • An AI agent that automates a repetitive business task
  • Augmented search over a document corpus
  • Deploying a model in-house to keep your data

Stack & tools

  • Python
  • LangChain
  • Hugging Face
  • PyTorch
  • Ollama
  • llama.cpp
  • RAG
  • MCP
  • CrewAI
  • TypeScript
/03 DIAGNOSTIC · ROADMAP

Data / AI Audit & Scoping

Before you invest, know where you stand. An outside view of your data and AI maturity, the concrete opportunities and the risks — so you leave with a clear, prioritised roadmap rather than a vague intention.

Who it’s for

Leaders and teams who want to know where to start, prioritise their data/AI work, or frame a project before launching it. Organisations mindful of their GDPR and AI Act obligations.

Methodology

Interviews and a review of what exists, a clear readout, optional support for delivery. No invented figures: whatever is stated is explained and justified.

Deliverables

  • A picture of your data and AI maturity
  • Prioritised opportunities (value × feasibility)
  • A short- and medium-term roadmap
  • GDPR and AI Act points of attention
  • Sobriety and cost-control leads

Typical use cases

  • Knowing where to start with data or AI
  • Prioritising work and framing a budget
  • Assessing a project’s GDPR / AI Act exposure
  • Setting a realistic roadmap

Stack & tools

  • GDPR (EU 2016/679)
  • AI Act (EU 2024/1689)
  • NIS 2 Directive
  • Data & AI good practices
/04 BUSINESS · USES · AUTOMATION

AI Training & Enablement

Helping your teams use AI day to day. Business-oriented training — understanding AI, using it concretely, automating your tasks — designed for non-technical users, not for ML engineers.

Who it’s for

Business teams, managers and leaders who want to demystify AI and gain autonomy: uses, prompts, copilots, task automation. Ideal to kick-start adoption in an organisation.

Methodology

A programme calibrated to your context and real cases. On-site in Brussels or remote, short and concrete formats. We start from your everyday tasks, not generic slides.

Deliverables

  • Leadership awareness — stakes, opportunities, AI Act
  • AI-usage workshops for business teams
  • Getting to grips with copilots — good reflexes (and limits)
  • Automating concrete everyday tasks
  • Support and resources to embed what’s learned

Typical use cases

  • Onboarding a business team onto generative AI
  • Learning to automate repetitive tasks
  • Framing a responsible AI use (AI Act basics)
  • Demystifying AI for a leadership team

Stack & tools

  • Brussels on-site
  • Remote
  • Real-case workshops
  • Business-oriented
  • No-code / low-code automation
/05 WORKFLOW · AGENTS · NO-CODE

Automation & Agents

Automating repetitive, time-consuming tasks — through workflows, AI agents and orchestration. Freeing up time for what matters, automating only where it is measurably useful.

Who it’s for

Operations, finance, support or HR teams losing time on manual processing, and scale-ups that want to make their processes reliable without adding headcount.

Methodology

Identify the automation candidates (effort × impact), prototype, roll out progressively. A frugal approach: no automation for its own sake.

Deliverables

  • Identification of tasks worth automating
  • Automation and orchestration workflows
  • AI agents connected to your tools
  • Documentation of the flows put in place
  • Support for hand-over and adoption

Typical use cases

  • Automating a manual, repetitive processing chain
  • Connecting tools that don’t talk to each other
  • Setting up an AI agent for triage or pre-processing
  • Making a reconciliation or re-keying reliable

Stack & tools

  • Power Automate
  • n8n
  • Zapier
  • Apache Airflow
  • LangChain agents
  • CrewAI
  • Python
  • TypeScript
/06 FINOPS · GREEN IT · FRUGALITY

Cloud, Cost & Sobriety

More sober, better-controlled cloud architectures. Cost analysis (FinOps), footprint-reduction levers (Green IT) and frugality by default — performance without the waste.

Who it’s for

Organisations whose cloud costs are rising, teams that want to control their bill, and leaders mindful of digital sobriety and their ESG goals.

Methodology

Analysis of costs and usage, right-sizing recommendations, identification of sobriety levers. We estimate footprint and prioritise the optimisations that are genuinely useful.

Deliverables

  • Cloud cost analysis and right-sizing recommendations
  • Footprint estimation and reduction levers
  • Frugality and sobriety good practices
  • Cloud governance recommendations (tags, budgets, alerts)
  • Sovereignty and hosting options suited to your needs

Typical use cases

  • Reining in a cloud bill that’s getting out of hand
  • Estimating and reducing a platform’s footprint
  • Setting up cost governance
  • Choosing suitable hosting (sovereignty, cost, need)

Stack & tools

  • Terraform
  • Kubernetes
  • AWS Cost Explorer
  • GCP Billing
  • Cloud Carbon Footprint
  • Docker
FAQ

Frequently asked

Answers to the most common questions — timelines, sectors, compliance, hosting, methodology.

Is Osyna a team or one person?

Osyna is an independent startup run by one person — Irvin Heslan, a data engineer based in Brussels. For larger needs I can draw on a trusted network. So you talk directly to the person who does the work.

Do you work with SMEs as well as large enterprises?

Yes. My background is in large enterprises (banking, energy), but the approach is modular and fits an SME or scale-up just as well. We start small, on a concrete need, and industrialise if it makes sense.

Which sectors do you know?

Mainly banking and finance, energy, industry, retail and professional services — the sectors I have actually worked in. For a sector I know less well, I say so, and we move forward together.

Do you take GDPR and the AI Act into account?

Yes, as good practices built into every project: minimisation, suitable hosting, AI Act points of attention. I am not a legal-compliance firm and do not certify a system: I build these requirements into the design and flag what should be taken further with a specialist.

Where is data hosted?

I favour European hosting and, where relevant, national providers, as well as self-hosting so you keep control of your data. It is not an absolute constraint: depending on your need, other options (including non-EU clouds) remain possible.

How long does an engagement last?

It depends entirely on the need — from a few days for scoping to several weeks for a rollout. I prefer short steps with concrete deliverables over announcing fixed timelines that mean nothing.

Which tools and languages do you use?

Mostly Python (data, AI, automation), SQL, and TypeScript for the web. On data: dbt, Airflow, Kestra, Databricks, BigQuery, Great Expectations. On AI: RAG, agents, LangChain, Hugging Face, Ollama, llama.cpp. I adapt the stack to your environment. Rust only if a very specific low-level optimisation warrants it.

Do you help my teams become autonomous?

Yes — it’s a goal, especially for business teams, junior profiles and management. The aim is that you can run and evolve what’s put in place without depending on me permanently.

Ready to unlock
your data?

Let’s discuss your transformation challenges. An initial conversation, no commitment, to frame your need and see how to move forward.