Gökhan Çetinkaya — Applied ML & AI Consultant

Who I work best with

You are probably a technical leader — CTO, VP Engineering, Head of Data — at a company that has tried ML before. Maybe you shipped something that worked in a notebook but stalled in production. Maybe you hired a team that delivered a dashboard nobody acts on. Maybe you are now thinking about LLMs and want someone who will tell you the truth about what they can and cannot do.

I work with people who want a rigorous partner, not a vendor who executes tickets. If you need someone to build a feature in a sprint, I am not the right fit. If you need someone to help you think clearly about a high-stakes ML problem and then help you execute it — that is what I do.

How I work

• Strategy & problem framing. Clarifying where ML can realistically move the needle — and where it is the wrong tool entirely.
• Solution design & prototyping. Translating a business problem into a data and ML architecture, and building something stakeholders can test and challenge.
• Production-oriented modeling. Classical ML, LLMs, agentic systems — with real attention to data quality, reliability, and what happens six months after deployment.
• Advisory & team mentoring. Helping technical teams think more clearly about metrics, experimentation, and when to stop optimizing a model and start shipping it.

Writings

Essays on applied ML, uncertainty, and decision-making.

• What Your Model's Mistakes Are Trying to Tell You
  The mistakes your model makes are one of the richest sources of signal in an ML project — and most teams never read them.
• What Is Correlation, Really?
  A word that means something precise in statistics and something much vaguer in most business conversations — and why the gap matters.
• Why Your AI Project Failed (and It Wasn't the Model)
  The hard part isn't getting your model to production. The hard part is keeping it useful once it's there.
• LLMs as Glue, Not Brains
  Why the organizations getting real value from LLMs are using them as connective tissue — not as reasoning engines.
• From Models to Decisions: A Practical Mental Model for Applied ML
  A conceptual framework separating prediction, uncertainty, dynamics, and decisions in real-world ML systems.
• Simulation: The Missing Layer Between Models and Decisions
  Why feedback loops, delays, and tail risk often dominate outcomes — and why simulation belongs between models and real decisions.
• Why Uncertainty Matters More Than Accuracy
  A practical essay on why point forecasts are often insufficient, and how uncertainty reshapes real business decisions.
• Why Accuracy Is Not Enough
  Why many ML projects fail at impact: model metrics aren't business KPIs, and point forecasts aren't decisions.
• Statistical / Bayesian Inference vs. Machine Learning: Rivals or Teammates?
  The real question is what you trust more — your model structure or your data.