I'm Igor Tarasenko, a product-minded engineer based in Amsterdam. I ship AI features to production with evals, guardrails, and monitoring, plus the handover-ready code your team can own.
What I Work On
Complete Solutions — I build end-to-end. iOS apps (WhisperBoard: 50k+ downloads, 4.8★), backends (FastAPI, Node, Rails), ML model deployment (CoreML, ONNX, cloud inference), web apps (React, Astro). Prototype to PRs to production.
Finding Real Leverage — I help teams identify where technology creates actual advantage vs where it's hype. Recent work: reduced a team's LLM costs from €2k/month to €40/month while improving quality. Cut another team's iOS build times by 35% through architecture changes.
AI Integration & Deployment — Pragmatic AI that works in production. On-device ML (CoreML, Whisper), LLM orchestration, evals, guardrails, and monitoring. I've been hands-on since SD 1.5, preferring direct, robust solutions over bloated frameworks.
Complex Architectures — 16 years dealing with messy systems. iOS (TCA, SwiftUI, modularization), backends (API design, data models, scaling), infrastructure (deployments, monitoring, CI/CD). Making large systems maintainable and teams productive.
Refactoring & Modernization — Incremental rewrites without breaking production. Mobile (UIKit → SwiftUI, MVVM → TCA), backend (monolith → services), infrastructure (local → cloud). Strangler-fig patterns, dependency injection, metrics-driven migration.
Developer Experience — Build tooling (Tuist, Gradle, Webpack), CI/CD pipelines, local development speed, onboarding documentation, debugging workflows. The stuff that makes teams ship faster or grind to a halt.
Hands-on GenAI Expertise
I've been deep in Generative AI since early 2023. I went to the first AI hackathon in Amsterdam and built an iOS app that changed room interiors using Stable Diffusion 1.5 hosted on my own Macbook.
Since then, I've been hands-on in the space:
- Early Adopter: I started using ComfyUI when it had less than 1,000 GitHub stars and have deployed SD and Flux models on platforms like Modal.com.
- Agentic Engineering: I've been automating my own software engineering work since GPT-4's release, using Python libraries for vector stores, structured output, and agents.
- Pragmatic Tooling: I was an early user of LangChain, but I've since moved away from it, as I find it's often overkill and promotes bad code. I prefer a direct, pragmatic approach to building.
Proof
- Shipped WhisperBoard — on-device transcription (50k+ downloads, 4.8★ App Store, 888★ GitHub).
- 35% build time reduction for large SwiftUI codebase via modularization and Tuist.
- 95% LLM cost savings (€2k → €40/month) through caching, routing, and prompt optimization.
- Team enablement: mentored engineering leads rolling out AI-powered workflows across mobile + backend squads.
Values
Ukrainian roots, Amsterdam base. I bias toward clarity, tight feedback loops, and making sure teams can operate without me after each engagement.
- Hands-on: I deliver diffs, docs, and recorded walkthroughs—not just slide decks.
- Transparent: One fixed-fee engagement with clear scope, deliverables, and handoff.
- Pragmatic: Every recommendation is grounded in production constraints.
How to Work Together
I have limited availability and typically work with 1–2 clients at a time.
My focus is on shipping production-ready AI. I offer one engagement: the AI Shipping Consultancy.