Przemysław Zalewski

AI‑native senior full‑stack engineer · from 3 November 2026

A senior engineer who makes AI-built software safe to ship and measure

Your team adopted Claude Code or Copilot, and now the review queue grows and quality slips? That gap is my job. Fifteen years of shipping software, nine in TypeScript, React and Node.js. In 2026 about two thirds of merged pull requests across two products I work on were built with coding agents, and nearly all went through my review or merge.

Book a free 30‑min call Email me instead

Calls this week · work starts 3 November 2026 · Wrocław, Poland (CET), remote

New: Quote Desk, from a customer’s inquiry to a draft quote, free

Przemysław Zalewski

The stack I ship with

  • TypeScript
  • JavaScript
  • React
  • Next.js
  • Node.js
  • Cloudflare
  • PostgreSQL
  • PostHog
  • Electron
  • Docker
  • GitHub
  • Anthropic
  • Stripe
  • Expo
  • .NET
  • MCP

Does this sound familiar?

Problems companies hire me to fix in 2026

Dependencies full of known vulnerabilities after npm audit?

On 14 October 2026, ten months after React2Shell, React published another advisory for unauthenticated remote code execution through React Server Components, rated CVSS 9.5, fixed only in 19.0.9, 19.1.10, 19.2.9 and 19.3.1. The longer upgrades wait, the bigger the jump when one is urgent.

I upgrade dependencies in waves you can revert one by one, a typical app often in a single evening. When React2Shell landed, I patched client apps the same day. In October 2026 I took a product from 313 vulnerable npm package entries, 15 of them critical, to zero across all 42 of its packages, and made the weekly scan fail on any new advisory.

Upgrades and CI/CD

Codebase still in plain JavaScript?

GitHub’s January 2026 post on typed languages says AI is pushing teams toward types, because more of the code now comes from agents and types are the safety net for code nobody wrote by hand. TypeScript is now the most used language on GitHub. Without types, an agent’s wrong guess about your data reaches production instead of failing the build.

I migrate to TypeScript with AI coding agents under enforced rules, file by file, with behavior pinned by tests before each step. Nine years of TypeScript tell me where strict types pay off first. In October 2019 I moved a shared JavaScript library of about 200 modules to TypeScript, keeping generated JavaScript for existing imports, and turned on strict mode three months later.

TypeScript migration

Legacy code nobody understands and no documentation?

Anthropic’s February 2026 post on legacy modernization calls understanding what the code does the most expensive step, because the authors are gone and the docs fell behind. Until someone maps it, every change is a bet.

An agent maps the code, I check the map, then I write the runbook and architecture docs from it and refactor behind tests, with a human review on every merge. In October 2026 I moved six legacy databases from MongoDB 4.2 to 9.0, through every major version in between, with 22 minutes of downtime and no rollback, then wrote the runbook so the next upgrade does not depend on me.

Legacy and custom software

Security questionnaires and assessments taking months?

Apps that use Google’s restricted scopes must pass a security assessment every year to keep access. Enterprise buyers add their own questionnaires, and the deal waits.

I led Google’s CASA Tier 2 assessment twice: passed in January 2026 and again at the yearly reassessment in October 2026. On that product the controls run as Playwright tests on every pull request, so the yearly review starts from test results instead of a blank form.

SaaS and compliance

Proof

Three things I shipped with agents in the loop

Problem, what I did, result. Client names only where I am allowed to use them.

Patriot Mail: an AI email client through CASA Tier 2 and onto the App Store

The problem
A greenfield AI‑native email client for web, iOS and macOS, built at Sanddev, needed Google’s CASA Tier 2 security assessment before it could read Gmail, and it had to work for older people.
What I did
I was the principal engineer at Sanddev. I wrote the delivery plan, shaped the product vision and treated prompt injection through untrusted email as a design constraint before any model touched a mailbox. CASA and OWASP ASVS controls became a Playwright suite that runs on every pull request. WCAG 2.2 AA sits behind an automated axe gate.
Result
I led CASA Tier 2 to a pass in January 2026, with third-party pentest findings fixed, and led the yearly reassessment, validated in October 2026; the questionnaire turned into a check on every PR. Live on the App Store on iOS and macOS since April 2026. Text scaling to 200% and a calm‑email mode shipped for older users. At that reassessment five items came back for missing evidence; I fixed the code first, then resubmitted updated evidence, and all five were accepted on 27 October.

CASA Tier 2 iOS + macOS WCAG 2.2 AA Open ↗

See also: SaaS & MVP · Mobile & desktop apps

Offer: From an AI‑built MVP to a product people pay for

An AI coding agent as a team member, under enforced rules

The problem
Coding agents write code faster than a team can review it. A team of four wanted the speed without a slop swamp on main.
What I did
I authored the rule set shared by Claude Code and Codex: mechanical checks over written rules, lint and test gates for security, API surface and CSP, agents working on main with hooks, every PR reviewed by a human, AI‑assisted content disclosed as such.
Result
In 2026 about two thirds of merged pull requests across two products were built with coding agents, and nearly all went through my review or merge. I merged 260+ PRs of my own and reviewed 300+ of other engineers’. My commits across those two products went from under 1,000 in 2025 to 7,500+ by October 2026.

~2/3 PRs with agents 300+ reviews 7,500+ commits in 2026

See also: AI & agents · Full‑stack & custom software

Offer: AI engineer for your team (Node.js, LLM agents)

An MCP server from first merge to first external users

The problem
Users of Sandtime.io, the time tracker I built at Sanddev, used by 4,000+ freelancers and teams, wanted to log time from their AI assistants. The API existed, the agent surface did not.
What I did
I set the tool surface, removed destructive tools, routed every operation through the existing REST controllers so an agent gets no wider access than a person, and instrumented adoption per client in PostHog from day one.
Result
First external users. A bot-protection rule that rejected MCP clients by user agent was found and removed the same day because the dashboard showed it.

MCP OAuth adoption dashboard

See also: AI & agents · PostHog & product analytics

Offer: Your company systems inside Claude and ChatGPT (MCP)

How we can work together

From a one-week audit to months inside your team

Ranges from my configurator at $36⁠–⁠54/h (€32⁠–⁠48) excl. VAT, base $45. See pricing. The final scope and price come in writing within 24 hours after a free call.

Best value for months of work

Dedicated contract: full capacity

$7,500 a month, net

€6,700 · 30,000 zł

About 167 hours a month (2,008 a year) at the base rate, with no surcharge for complex work.

  • My working weeks go to your product first, with priority in my calendar.
  • A standing channel (Slack or similar) and a reply the same working day.
  • Complex work at the same price: recovery, payments, security, legacy.
  • Three months minimum, then one month’s notice on either side.
  • A B2B services contract: the scope of services in writing, my own tools and way of working, one invoice a month in USD, EUR or PLN.

Book a contract call → Set the term in the configurator Details and rules on the pricing page →

1 week

One-week code audit

I read your code, tests, dependencies, deploys and analytics by hand, plus your API and auth if agents or MCP are on the roadmap. You get a written report: risks ranked by cost, what to fix first, what an agent could safely do today, a plan with milestones.

  • Findings you can act on
  • A plan with milestones
  • A go or no‑go you can defend

$1,080⁠–⁠$2,160 excl. VAT · 30⁠–⁠40 h

See price and start date →Book a free call

3 to 6 weeks

AI feature or rescue sprint

One outcome, shipped and measured: an MCP server, an AI feature with an eval loop, a vibe‑coded MVP made production-ready, a Next.js app that stops hurting.

  • Something to touch in week one
  • Tests around what matters
  • A tracking plan and dashboard

$3,672⁠–⁠$7,452 excl. VAT for 3 weeks at full capacity · 102⁠–⁠138 h

See price and start date →Book a free call

2 to 3 days a week

Fractional tech lead

Inside your team for at least three months: architecture, code review, the rules for coding agents, hiring screens, the estimates. Your team keeps shipping, I keep the direction and the quality bar.

  • Code review for the team
  • Agents under rules
  • Numbers at the weekly call

$6,732⁠–⁠$12,366 excl. VAT for 2 days a week over 3 months · 187⁠–⁠229 h

See price and start date →Book a free call

Need someone for the long haul? The dedicated contract above is how I prefer to work: as an AI engineer, Forward Deployed Engineer or AI enablement lead, from November 2026. Profile and terms →

Or start from a specific problem

Services

Pick the technology. Each page has a live demo.

Front-end

React

Interfaces, tests and the dashboard that proves they work.

Live demo: editable component with Profiler →

Next.js

App Router done right, or an honest exit plan.

Live demo: waterfall vs stream timeline →

TypeScript & Node.js

Typed APIs, Postgres at scale, tests that encode security.

Live demo: JSON to TypeScript types →

JavaScript

Browser and Node.js, with or without a framework, legacy included.

Live demo: the event loop, visualized →

Platform, delivery & product

Cloudflare

Workers, Durable Objects, R2, Workers AI, with cost caps.

Live demo: served from the edge, live →

Maintenance & CI/CD

Upgrades with a safety net, migrations, GitHub Actions, legacy kept alive.

Live demo: dependency upgrade planner →

Full‑stack & custom software

Scope to production, solo or as fractional tech lead.

Live demo: estimate with an AI coefficient →

SaaS & MVP

From first commit to paying users: billing, tenants, stores.

Live demo: first four weeks planner →

Mobile & desktop apps

React Native, Electron, PWA, every store that matters.

Live demo: your device, probed live →

Email

IMAP clients, templates that survive Outlook, SPF, DKIM, DMARC, BIMI.

Live demo: check your domain’s email posture →

AI, data & specialties

Agentic engineering

Give me the goal, I ship it end to end, with coding agents by default.

Live demo: one real task, goal to production →

AI in your company

Adoption plans, custom agents, MCP, skills and rules for coding agents.

Live demo: AI readiness scorecard →

AI & agents

Agents in production: tools, MCP, evals, cost caps.

Live demo: ask my agent about me →

PostHog & product analytics

Tracking plans, first‑party proxy, dashboards leadership reads.

Live demo: live event feed of this page →

C# & .NET (legacy)

Keep the old system alive, then move it without a rewrite.

Live demo: C# to TypeScript translator →

Games

Browser, desktop and mobile games, from the loop to the leaderboard.

Live demo: thirty seconds of arcade →

How I work with AI

AI-native means rules, review and numbers

Rules the agents cannot skip

Mechanical checks over written guidelines: lint and test gates for security, API surface and CSP. The model suggests, the criteria decide.

A human answers for every merge

Agents open pull requests, I decide what merges. The gates catch the mechanical problems, I review the rest, and the responsibility stays with me.

Quality measured, never assumed

Prompts and models are scored against saved test cases: best‑of‑n, LLM‑as‑judge. I have compared many models across releases and swap them when the scores say so.

Disclosed, so trust survives

AI‑assisted work is labeled as such, to the team and to users where it matters. Agents are teammates, not ghostwriters.

~2/3of merged PRs in 2026 built with AI coding agents, nearly all through my review or merge
260+merged PRs of my own in 2026, plus 300+ reviews of other engineers’ PRs
270+tagged releases of one product in 10 months of 2026
~3×more GitHub contributions in 2026 than in 2025, 8,500+ by October, with agents under enforced rules

Free tools

Use them without talking to me

Small tools I built and run on Cloudflare. No account, no email gate.

Quote Desk

A customer inquiry (email, PDF, Excel or a scan) matched to your price list, unclear lines flagged, exported to Excel.

Open →

AI Visibility Desk

Does AI recommend your brand? Your buyers’ questions asked every week with web search on: who is named, who is cited, the trend.

Open →

Vitals Desk

Core Web Vitals checked every day on a throttled phone: LCP, CLS and TBT, a screenshot, the change since yesterday and a badge for your site.

Open →

List Desk

An email list checked in your browser: dead domains, null MX, disposable providers, typos and duplicates. Addresses never leave the page.

Open →

Eval Desk

Test cases for an AI assistant, run after every prompt change: what regressed, what got fixed, which cases are unstable, and the prompt diff.

Open →

Dashboard Desk

A growth dashboard created in your own PostHog: channels incl. AI assistants, funnel, activation, retention, errors, plus what your tracking is missing.

Open →

Signature Desk

Company email signatures for the whole team: one template, the legal footer, Outlook‑safe HTML for Gmail, Outlook and Apple Mail.

Open →

All 25 free tools →

Press

Written up elsewhere

Real links, from Windows Phone apps in 2012 to a game in 2020. Online I go by Ciunkos. Some outlets are gone, so those links open an archived copy.

Antyweb Award“The jury picked the best apps in the AppAward contest”AppAward 2013: Pogaduszki won the Other category on Windows Phone, out of 300 apps entered across iOS, Android and Windows Phone.Read (in Polish) Microsoft Polska“The best Polish apps for Windows Phone”Microsoft Poland’s hand‑picked list of the best Polish Windows Phone apps featured my Kwejk app (archived copy from September 2012).Read the archived copy (in Polish) wp7.com.pl“More about the Gadu app”A chat with me about Gadu: “for that period Gadu was the most downloaded app on the Polish Marketplace.”Read (in Polish) Antyradio“Fight the epidemic as the government and win the Korona Krulów. A satirical game with TVP news bars”“The work of one programmer, who built it literally at night.” Antyradio’s review of the game.Read the archived copy (in Polish) Gazeta Plus“The game »Korona Krulów«. Lead the PiS government and make decisions in a time of plague…”News piece on the game, the day after launch.Read the archived copy (in Polish) Wieści24“…A game about the coronavirus in Poland has appeared online: Korona Krulów”News coverage of the game the day after launch.Read (in Polish)
More mentions (9)
Wykop“Feel like the Chairman and run the country in Korona Krulów: the game is out”Korona Krulów launch day: 698 upvotes and 152 comments.Read (in Polish) OnMSFT“Windows Store app highlight: Windows Phone Store, grab the latest Windows Phone apps via Windows 8”My Windows 8 app for browsing the Windows Phone Store, featured as a highlight.Read windows‑phone.pl“The »Windows Phone Store« app for Windows 8”“Built by Ciunkos, known for apps like Pogaduszki and Memgenerator.”Read the archived copy (in Polish) appsblog.pl“Memgenerator: make a funny picture on your phone”Review of Memgenerator, my Windows Phone meme maker.Read (in Polish) Komórkomania“Pogaduszki: a Gadu‑Gadu client for Windows Phone”Review of Pogaduszki, my Gadu‑Gadu messenger: 4.5 stars from 68 ratings at the time.Read (in Polish) Telix.pl“Pogaduszki: a new GG client for Windows Phone”Launch of Pogaduszki for Windows Phone 7.5: free, no ads, push through my own server.Read (in Polish) wp7.com.pl“Unofficial GG client »Pogaduszki« now available”Pogaduszki ready “after many hours of work and testing”, no splash screen so it starts faster.Read (in Polish) windows‑phone.pl“Pogaduszki, a new GG client!”After my first Gadu‑Gadu app was pulled over a copyright claim, I rebuilt it as Pogaduszki without the infringing parts.Read the archived copy (in Polish) GaduNews“GG Network pulls an unofficial messenger from Windows Phone”Gadu, the first native Gadu‑Gadu client on Windows Phone 7, comes down after a trademark claim. Quotes my statement.Read the archived copy (in Polish)

One foot in the door

Not ready for a project? Start with one pull request.

A line‑level review of the change that worries you, before you commit to anything bigger. The one‑week code audit is in the cards above.

Code review of one pull request or one module

Send me the diff that worries you. Within a few days you get line‑level comments, the three things I would change first and whether I would merge it. I have reviewed 800+ pull requests on one product since 2018.

$72⁠–⁠$216 excl. VAT · 2⁠–⁠4 h · See price and start date →

Book a call Free 30‑minute call first. If it does not fit, I say so on the call.

How it works

Three steps from this page to a running project.

  1. 1

    A 30-minute call

    You describe the product and the problem. I ask the questions I would ask on day one.

  2. 2

    Scope in writing

    Within 24 hours after the call you get a short written plan: what we build first, what we measure, how we work together and what it costs. No slide deck.

  3. 3

    Working software in week one

    Deployed, measured and open for your review. After that, the numbers set the pace.

Book a free call Email me instead

Fit check

Is this a fit?

A good fit when

  • You need a senior who owns delivery, not a pair of hands
  • The product has to ship and be measured, with or without a product manager
  • A solo engagement, or one team of up to four that I lead directly
  • TypeScript, React, Node.js, Cloudflare, Postgres, PostHog, and Python when the work needs it

Probably not when

  • A program that needs a ten‑person vendor team
  • A stack built around Java or a cloud I do not work in
  • Body leasing by the hour with no say in how the work is done

FAQ

Questions people ask before they book

What does AI‑native mean in practice?
Coding agents are team members with rules, not autocomplete. I write the rule set they work under, keep lint and test gates in CI, review what they open and measure the result in PostHog. The AI and agents page shows the agent that runs on this site, with its tool calls visible.
Can you take over a vibe‑coded codebase?
Yes. If the app is stuck before launch, start with a rescue: 2 to 8 hours, $72 to $432 excl. VAT. If it has paying users and keeps breaking, first a one‑week audit with findings you can act on, then tests around the flows that make money, then the three biggest risks. The site keeps running throughout. The full‑stack page describes the take‑over process and the Next.js page has a live demo of the most common performance problem.
Remote, time zones, language?
Remote from Wrocław, Poland (CET). I have worked across a nine‑hour gap with a US client and daily with UK teams. English is my working language, Polish is native. This site exists in both.
Teams or solo?
Both. Solo for audits and sprints. As fractional tech lead I lead up to four engineers directly, plus their agents. The agent rules and CI gates land in your repositories, so every team that commits there works under them, and the team workshop runs on your own repos. If you need ten more people, I will tell you on the call and point you to a team.
You are one person. What if you get sick or leave?
The work does not leave with me. Code, CI and docs live in your repository and accounts from day one, every project keeps a runbook and architecture notes written from the code, and notice is one month on both sides with a written handover. If I am ill, you hear it the same day and the plan moves, not the quality. How I work.
Why not an agency or a software house?
You talk to the person who writes the code: no account manager, no salesperson, no bench, no minimum project. The dedicated contract buys about 167 hours of senior work a month; an agency invoice for one developer also pays for its project manager, sales and margin. If you need ten people, I will say so on the call.
We already have Claude Code or Copilot. Why hire you?
Because the bottleneck moved from typing to review, rules and judgment. In 2026 about two thirds of merged PRs across two products I work on were built with coding agents, and nearly all went through my review or merge: 260+ PRs of my own and 300+ reviews of other engineers’. I set up the same rules and CI gates for your team, as pull requests in your repositories.
Two thirds of PRs from agents. Who actually reviews them?
A person, and on my projects that is me. CI gates reject the mechanical problems before review, so my time goes to the product logic and the reason for the change. I have reviewed 800+ pull requests on one product since 2018, and a public LinkedIn recommendation says: “I always find his code reviews especially valuable”.
Who owns the code?
You do, from day one: the code in your repository, the domain, hosting and every account in your name, and the contract transfers the copyright to you in writing. I never hold the last part of a project hostage until you sign a retainer. No proprietary framework, no lock‑in: you can stop any day and keep everything.
How can I check you before a call?
The Polish business register (CEIDG), tax ID and D‑U‑N‑S number are on the trust page, next to public LinkedIn recommendations, and the fixes I got merged into widely used libraries are on the open source page. Or ask any AI assistant to do due diligence on me: everything on this site is public and checkable.
An agency sent us a quote. Can you check it?
Yes. 4 to 8 hours, $180⁠–⁠360 excl. VAT: I go through the estimate line by line, the risks nobody priced, what can leave the MVP and whether the team plan matches the scope. You get it in writing, and you decide what to do with it.
Contract and rates?
Preferably a dedicated B2B contract at full capacity: $7,500, €6,700 or 30,000 zł net a month, about 167 hours, three months minimum. Hourly work with no minimum is fine too. From 3 November 2026. Invoices in USD, EUR or PLN, whichever is easier for you. Base rate $45, €40 or 180 zł an hour net; simple work 20% less, complex 20% more. Details on pricing. The final scope and price come in writing within 24 hours after the first call.
What if it does not work out?
Start with the two‑week trial: five tasks from your backlog, price fixed before we start, code in your repo. Then you decide. Hourly work has no minimum either: stop at any time and keep everything done so far, because the code is in your repository from day one.
Does a reply within 24 hours mean only an acknowledgement?
No. On working days you get a full answer, and for a project issue often a pull request or a fix. Weekend messages get an answer by Monday noon; weekend work only by agreement, at twice the rate. No SLA and no on‑call. The rules are on how I work.
Who sees my code and data, and do you use subcontractors?
Only me. No subcontractors. AI tools run in the mode you pick, with training off, and secrets, customer data and logs never reach a model without your written approval. Facts for your vendor questionnaire are in the trust center.
How soon can you start?
Work starts 3 November 2026; calls happen now. Book the free call and the written scope lands within 24 hours after it. My resume has the timeline.

Bonus · #suchar

A dad joke, on the house

Free, like the first call. Same sense of humor, included in every project.

How many programmers does it take to change a light bulb?Punchline

None. That’s a hardware problem. Ba‑dum‑tss.

Why do programmers mix up Halloween and Christmas?Punchline

Because OCT 31 = DEC 25. Ba‑dum‑tss.

Why did the developer go broke?Punchline

He used up all his cache. Ba‑dum‑tss.

In 30 minutes you will know if we are a fit

You bring the problem, I bring the questions. Want a number first? The configurator shows hours, a price range and the start date.

Book a free 30‑min call Email me instead

or write to ciunkos@gmail.com

After you book you get a calendar invite. Nothing else to fill in.

My pledge

  • Reply within 24 hours. I reply within 24 hours on working days, personally, not a bot. Weekend messages: by Monday noon.
  • Late means free. If I am late, the first hour of work is free. You do not have to ask.
  • Scope in writing within 24 hours. A free 30‑minute call with no company deck, then the scope and the price in writing within 24 hours. You decide.
  • The person who does the work. You talk to the person who does the work. No sales layer, no handoff.
  • Follow‑up only with news. I follow up only when I have something new for you. Never a “did you see my offer”.
  • Honest “not a fit”. If I am not the right person, I say so on the call and point you to someone who is.

Calls this week · work starts 3 November 2026 · B2B contract · Wrocław, Poland (CET), remote

You own the code, domain and accounts from day one. Nothing is held hostage.