Work
Built, shipped,
and still running.
Selected work across GTM systems, AI tooling, and product delivery. Grouped by what it proves, not by when it happened.
Cluster 01
AI-native GTM systems
Practical AI implementation, workflow redesign, and code-assisted execution. Turning ideas into working operator tools.
AdsAI / Ad Assistant
Google Ads cockpit with AI-driven ICP lead scoring (0-100). Shifts optimisation from CPL to CPL-ICP-fit, with campaign priorities and next-best actions in one workspace.
- Problem
- Ad optimization work was split between reports, intuition, and repeated checks.
- System
- Budget, ICP, ROAS, alerts, and actions are pulled into one decision loop.
- Artifact
- Live dashboard with campaign priorities and optimization queue.
Claude Code GTM Agent Starter Pack
Open-source starter pack for building AI GTM agents with Claude Code. Designed for growth operators who want to ship AI-native workflows without starting from scratch.
- Problem
- Operators wanted AI-native GTM workflows but had no structured starting point: just blank Claude sessions and scattered prompt tips.
- System
- 4-layer agent stack: CLAUDE.md operator identity → MCP tools → client memory files → 5 reusable operator skills. Covers research, enrichment, sequence drafting, and CRM updates.
- Artifact
- Open-source GitHub repo with full starter pack: CLAUDE.md template, skill definitions, folder structure, and MCP source.
NotchCue
Native macOS teleprompter that places private notes around the MacBook camera notch. Keeps eye contact during calls, demos and recordings.
- Problem
- Notes sit away from the camera, which breaks eye contact during calls and demos.
- System
- Prompts live around the MacBook notch with a focused reading surface.
- Artifact
- Product site and app concept built around the camera area.
Outbound intelligence system
Replaced a broken spreadsheet-and-gut-feel outbound motion with a structured AI agent stack: enrichment, scoring, and sequence drafting in a single operator loop.
- Problem
- A B2B SaaS team was running outbound from a mix of manually-enriched spreadsheets, Apollo exports, and instinct. Coverage was inconsistent, personalization was template-based, and the cycle from list to sent sequence was two to three weeks.
- System
- Claude Code with four MCP integrations (LinkedIn, Apollo, HubSpot, Clearbit), a client-specific CLAUDE.md operator brief, and named skills for each stage: account scoring, ICP classification, signal extraction, and sequence drafting. Each run produces scored accounts with drafted sequences ready for human review, not a bulk send.
- Artifact
- Operator stack: CLAUDE.md brief, four MCP connections, five named skills. A 50-account cold CSV goes to scored accounts with personalized sequence drafts in under an hour. Output reviewed and approved before any contact is made.
Wojciech Coach · Apple-native running coach
Private running and activity coach for iPhone, Apple Watch and Mac. HealthKit as the single source of truth, live workouts on Watch, adaptive training plan, GPS maps, badges and analysis. Built with SwiftUI and Codex.
- Problem
- Existing fitness apps either require subscriptions, send health data to remote servers, or fragment the experience across devices without a single coherent plan.
- System
- Apple-native stack: HealthKit reads all workouts and health signals, Apple Watch runs live sessions via HKWorkoutSession, iPhone synthesises the data into a daily decision and adaptive weekly plan, Mac provides the full analysis dashboard. All data stays on device.
- Artifact
- Functional MVP: iOS app, watchOS workout app, macOS analysis centre, WidgetKit extension with Live Activities, and a landing page.
Klaro · local-first Mac care
Native macOS utility that turns cleanup, storage checks and system reports into a calm review flow. One scan, one plan, decision always on the user's side.
- Problem
- Mac cleanup apps delete silently, use scare tactics, and give no audit trail. Users either ignore the problem or hand control to a black box.
- System
- SwiftUI dashboard with local scan engine, CLI for audit and duplicate detection, safe cleanup preview that moves files to Trash (never silent delete), and HTML report export with JSON-LD and LLM-readable metadata.
- Artifact
- Working MVP: macOS app bundle, CLI binary, and HTML report with structured metadata. Scan, preview, act: three-step loop that keeps the user in control at every stage.
Cluster 02
Growth architecture in practice
Revenue-system thinking with real metrics, operating cadence, and growth levers. Built and run, not just scoped.
Revenue architecture · Series A vertical SaaS
Rebuilt the commercial model for a vertical SaaS company approaching Series A: ICP definition, pipeline architecture, and a conversion system that reduced founder dependency.
- Problem
- The company had strong product-market fit in a tight vertical but a commercial model built for the founder's network, not for scale. Pipeline was unpredictable, ICP was broadly defined, and every deal above a threshold required founder involvement to close.
- System
- ICP redefinition scoped to the three highest-converting customer profiles. Pipeline stages rebuilt around buying signals rather than time-in-stage. A qualification framework that let AEs triage and advance deals independently. Conversion metrics instrumented for the first time to make improvement legible.
- Artifact
- ICP playbook, stage-gate qualification framework, deal velocity benchmarks by segment, and a handoff protocol that moved the founder from primary closer to deal sponsor on strategic accounts only.
Full-stack growth system · mid-market B2B SaaS
Built a unified revenue operating model from scratch for a growing B2B SaaS product. CRM architecture, SEO engine, CRO layer, winback automation and analytics stack: designed and shipped as one integrated system, not a collection of disconnected tools.
- Problem
- Marketing was a series of one-off campaigns with no operating model underneath. CRM held contact data but no lifecycle logic. SEO was ad hoc. Winback didn't exist. There was no single place that showed where revenue came from and what to do next.
- System
- Designed and shipped five interconnected layers as a single operating model: CRM with lifecycle stages and lead scoring, SEO architecture with content and technical foundations, CRO system with A/B testing and conversion tracking, winback automation for churned and inactive segments, and a unified analytics layer tying all of it to revenue outcomes.
- Artifact
- Dashboards, playbooks, automation workflows, CRM configuration, and a handoff operating manual. The system ran without me: that was the acceptance criterion.
Cluster 03
Products actually shipped
Real product delivery with UX and conversion thinking. Proof of range beyond strategic documents.
Relora
A relationship operating system for local-government property teams. One place for contacts, owner context, communication status, and next actions.
- Problem
- Important owner conversations lived across inboxes, notes, and memory.
- System
- Contacts, relationships, briefs, status, and follow-up in one operator view.
- Artifact
- Live demo app with dashboard, relation map, and task flow.
Camper Rental Booking Engine
Camper rental booking engine with dynamic pricing, availability management and a clean reservation flow.
- Problem
- Rental demand needed a clearer path from browsing to booking.
- System
- Inventory, dates, offer framing, and reservation CTA in one front-end flow.
- Artifact
- Responsive booking website with vehicle cards and search entry point.
Działka+, Paczka+, Resume+
A cluster of focused utility apps shipped independently: a plot management tool, a parcel tracking aggregator, and an AI-assisted resume builder. Each solves one real problem. Each is in production.
- Problem
- Recurring operational gaps that SaaS subscriptions were too heavy to fill: each problem needed a focused tool, not a platform.
- System
- Individual apps designed around one job-to-be-done each. Built with AI-assisted development, deployed independently, and running in production.
- Artifact
- Live apps available at app.wojciech.io/apps: plot management, parcel aggregation, AI resume builder.
HireMe
Job search command centre: funnel, priorities and activity in one view. Public case study coming after the first cohort of users.
- Problem
- Job search work fragments fast: jobs scattered across tabs, applications half-tracked in spreadsheets, no signal on priorities or pipeline health.
- System
- One dashboard for the whole funnel: saved jobs, applied, interview pipeline, offers, with an activity log and quick filters by priority and status.
- Artifact
- Private PWA live in the app workspace. Five-stage pipeline view, activity log, priority filters, deployed and in use.
Next step
Have a system worth fixing?
Send the context, the current bottleneck, and what already exists. I will tell you where I would start.