Oleksandr TverdokhliebAI as part of the toolchain
How I use AI day to day — to build faster, verify more, and get more out of data. This site itself was built with an AI coding agent.
AI isn't a side experiment in how I work — it's part of the toolchain. I use it deliberately across the whole delivery cycle: writing and refactoring code, reviewing and verifying it, and exploring and explaining data. The point isn't novelty; it's leverage — shipping reliable analytics products faster, with more checks, not fewer.
How I use it
Development & bug-fixing
As a pair-programmer — building features, refactoring, and tracking down bugs faster, while I keep ownership of the architecture and the decisions that matter.
Agentic automation
Point an AI agent at a codebase or a recurring data task and let it run the multi-step work — reading context, making changes, executing and checking its own output.
Exploring & analysing data
Profile new datasets, draft and debug SQL and pandas, surface patterns and outliers, and pressure-test a hypothesis before committing to it.
Extracting structured data
Turn messy PDFs, images and emails into clean, structured rows — like the leaflet-to-data pipeline behind Leaflet Analyzer.
Reporting & narrative
Turn numbers into decision-ready writing — executive summaries and "what changed and why" commentary, drafted by AI and edited by me.
Data quality & validation
Generate validation rules, flag anomalies and edge cases, and review changes for errors before anything reaches a stakeholder.
Forecasting & modelling
Draft baseline models, suggest features, and help interpret results and trade-offs — accelerating the analysis without outsourcing the judgement.
Talk to your data
Natural-language questions over a governed dataset — so business users can ask "why did the East region dip?" and get a grounded, checkable answer.
Classify & tag at scale
Categorise products, tickets and free-text feedback, and standardise messy labels — work that's slow by hand and well-suited to AI with a human check.
How I work with AI
Human in the loop
AI drafts; a person reviews and approves. Nothing ships, or reaches a stakeholder, without a human signing off.
Verify, don't trust
Every AI-generated change is tested and checked. The goal is more verification, not less — AI is a second set of eyes, not a shortcut around them.
I keep ownership
Architecture, modelling choices and the final call stay with me. AI accelerates the work; it doesn't make the decisions.
Auditable & governed
AI-assisted outputs are held to the same governance as the rest of the data work — traceable, reproducible and reviewable.
AI, ML & data-science courses
All learning →The certifications behind how I apply AI and machine learning to analytics — from foundations to applied ML and NLP.
Loading…