# Łukasz Tomaszewski — Engineering Manager & AI-First Software Engineer Engineering Manager · Mobile SDK & Ad Tech Building mobile SDKs reaching billions of impressions Latest Articles ## Articles ### I Spent 6 Months Building AI Tooling for a Problem Nobody Was Measuring. Then I Open-Sourced It. LinkedInApr 2026Open Source · AI Engineering · Claude Code · TypeScript Every AI workflow I ran was a black box — no cost visibility, no quality tracking, no way to compare approaches. So I built mcp-ts-engineer: spec-driven TDD, multi-agent orchestration, per-session cost tracking, and automated code review for Claude Code. It turned my guessing into engineering. The repo is public because this problem isn't just mine. [→ Read on LinkedIn](https://www.linkedin.com/in/ltomaszewskiconsulting) ### Half My Calendar Is Empty on Purpose. I Use It to Ship Code. LinkedInApr 2026Engineering Management · Leadership · Distributed Teams · Mobile SDK I lead a 4-engineer distributed squad building mobile SDKs and write code every day. The default EM calendar is packed with alignment meetings and status updates. I deleted half of mine — kept 1:1s, sprint planning, and incident response. Everything else has to earn its slot. The result: I catch architecture issues before standup, I pair on hard bugs without context-switching, and my team stopped asking 'can you review this?' — because I'm already in the codebase. [→ Read on LinkedIn](https://www.linkedin.com/in/ltomaszewskiconsulting) ### Same Feature, 3 Ways: Manual vs AI Agent vs Multi-Agent Pipeline LinkedInMar 2026AI · Multi-Agent Systems · Software Engineering · TypeScript The cheapest option wasn't the best. The fastest wasn't the cheapest. And the most expensive one had zero bugs. I built the same feature three ways — by hand (4 hours, $0), with a single Claude Code session (45 minutes, $1.80), and through a multi-agent pipeline with spec review, TDD, and automated auditing (20 minutes, $3.40). Here's what the data says about when each approach actually wins. [→ Read on LinkedIn](https://www.linkedin.com/in/ltomaszewskiconsulting) ### 3 Numbers I Track Every Time I Use Claude Code LinkedInMar 2026AI · Claude Code · Developer Productivity · Metrics AI makes some tasks 5x faster and others quietly 2x worse. The data shows which is which. For my first 50 sessions I had no idea if I was improving or just getting comfortable with mediocre output. Then I started tracking three metrics: cost per workflow, lines of code kept vs discarded, and fix cycle count. After 13 years of shipping in Swift and Kotlin, this is how I learned to ship with AI. [→ Read on LinkedIn](https://www.linkedin.com/in/ltomaszewskiconsulting) ### I Built My Portfolio 3 Times With AI Agents. The Third Cost 70% Less. LinkedInMar 2026AI · Claude Code · Build In Public · Cost Tracking The difference wasn't the AI model. It was what I measured between attempts. Three tries at building a portfolio site with Claude Code — first attempt: $4.50 and two hours debugging garbage. Second: $2.10 and noticeably better, but I couldn't explain why. Third: $1.40 and production-ready. Here's exactly what changed. [→ Read on LinkedIn](https://www.linkedin.com/in/ltomaszewskiconsulting) ## About Engineering Manager at Chartboost leading a distributed Mobile SDK squad. My team maintains and extends ad tech SDKs serving billions of impressions across iOS and Android. AI automation powered by Claude Code and Claude Agent SDK offloads the routine work so the team stays focused on what matters for the business. 13+ years shipping native mobile — started in Swift and Kotlin, now TypeScript is my primary language. Led teams from zero-to-one at a fintech challenger bank and built apps that secured Shark Tank funding. Outside work I explore new tech through side projects. Currently deep into Claude Code, agentic workflows, MCP servers, and multi-agent systems — proving how they multiply what small teams can deliver. Projects like [Mellow](https://apps.apple.com/us/app/mellowsleep/id6636552824), Bastion, Voice Note Taker and more. Based in Warsaw, Cracow, and San Francisco. ## Selected Work ### Chartboost Ad Tech · Swift SDK · Kotlin SDK · Claude Code · Claude Agent SDK Leading a distributed squad of 4 engineers developing Mobile SDKs and apps for a leading Ad Tech platform. Reduced operational overhead by 60% via AI automation. Directed greenfield development and store launch of Chartboost AdPreview, transforming IAB Certification into a self-serve workflow. [→ Live](https://loopme.com/chartboost/) ### Mellow Health Tech · React Native · Swift · Kotlin · Claude Code A sleep and wellness iOS app built as a side project. Designed and shipped end-to-end — from concept to App Store. Built with Swift and SwiftUI, exploring health-focused UX patterns and Apple ecosystem integrations. [→ App Store](https://apps.apple.com/us/app/mellowsleep/id6636552824) ### Overplay Games Mobile Games · Swift · Kotlin Multiplatform Led a remote mobile engineering team of 4 for a no-code game creation platform. Modernized the legacy iOS codebase to SwiftUI and architected a Kotlin Multiplatform solution for Android. Delivered UX enhancements that directly supported a successful Shark Tank pitch with Mark Cuban. [→ App Store](https://apps.apple.com/us/app/overplay-games/id1486242754)[→ Google Play](https://play.google.com/store/apps/details?id=com.overplaygames) ### Batmaid Ecommerce · Swift · iOS Architected and built the mobile application ecosystem from scratch for a leading on-demand cleaning services platform. Assembled and led the founding mobile engineering team, delivering a scalable iOS solution that drove 30% revenue growth. [→ App Store](https://apps.apple.com/pl/app/batmaid-home-cleaning/id1527257152)[→ Google Play](https://play.google.com/store/apps/details?id=com.batmaid.batmaid&hl=pl) ## Selected Open Source 2184 contributions in the last year. Apr 2026: 118 contributions across 8 active days. This week: 3 contributions (1 of 2 days active). ### MCP TypeScript Engineer Claude Agent SDK · MCP · TypeScript · Multi-Agent An MCP server that orchestrates multi-agent development workflows — spec-driven development, TDD implementation, code auditing, and PR review. Built with the Claude Agent SDK for autonomous engineering pipelines.