H2MHuczyńskiSystems you own
Contact

Services · H2M

Four problems I take off your plate

Each one ends with a system your team can operate without me, on your own accounts, with the code in your hands. Pick the one that sounds like your week.

Book a 30-minute callSee the work

  1. 01 / 04 · AI Engineering

    Stop paying per token for a task that has not changed: a small model on your own server

    Distill, fine-tune and deploy the smallest model that does the job, then wire it into the place your team already works.

    • Model distillation
    • Fine-tuning on your corpora
    • Quantized CPU deployment
    • Agents with MCP tool access
    • Eval suites with published negatives
    Full page

    The right call when

    1. Your frontier-API bill grows every month for a task that has not changed since spring.
    2. Your team already lives in Slack or Claude and wants the CRM, the CMS and the reports reachable from there.
    3. You need an agent that acts on company data, with every write audited.

    You get

    • Eval suite on your data, with the negatives published
    • The smallest model that passes it, quantized, deployed on CPU or your account
    • MCP server exposing your systems as tools, one write-gated gateway
    H2M ML Lab

    Thirteen distillation experiments on real corpora. Students from 230M to 1.2B parameters, quantized, about a second on CPU.

    See case study
    Adtrip

    Slack-native AI operator for a 25-person agency: headless Claude Code behind one write-gated gateway.

    Read the case
    Adcare

    A clinic marketing operating system operated from Claude through its own MCP server.

    Read the case

  2. 02 / 04 · Systems: architecture, build, operation

    One system instead of three vendors: site, CRM and automations on your accounts

    Web products, internal apps and e-commerce shipped as one codebase on your own Cloudflare account, and kept running after launch.

    • Web development
    • Internal apps
    • E-commerce and configurators
    • CRM and automation as one system
    • Fractional operation with proof of last run
    Full page

    The right call when

    1. The site, the CRM and the automations come from three vendors and nobody owns the whole.
    2. You need a product, an internal app or a configurator that has to be live before the quarter ends.
    3. Something is running and nobody can say whether it ran last night.

    You get

    • Architecture on one page, then the system on your Cloudflare account
    • Site, CRM, automations and the MCP server as one codebase
    • Scheduled tasks that leave proof of their last run
    Seecontainerpool

    A Swiss brand with a 3D pool configurator; one pricing truth shared by the configurator and a partner API.

    Read the case
    Adcare

    Multi-tenant CMS plus agency and per-clinic CRM for a clinic marketing agency.

    Read the case
    pozyskajpacjenta.pl

    Clinic SaaS shipped in under four weeks: custom hostnames, booking, CRM, Stripe. Zero paying users so far, published as-is.

    Read the case

  3. 03 / 04 · AI training for teams

    Your team automating its own work in a day, safely

    Workshops for business teams built from 75-minute modules, up to four a day, on site or remote. Bring any device with a browser: the sandbox, the tools and the API tokens come from me.

    • Entry module: no setup
    • AI agents in daily work
    • Git for everyone
    • Office documents from prompts
    • Your own MCP servers
    • Browser automation
    • Evals and guardrails
    Full page

    The right call when

    1. Your team pays for AI subscriptions and uses them like a search box.
    2. Two people in the company can automate things; the other twenty send them tickets.
    3. You want the team to build its own small tools and agents, safely, and keep them in git.

    You get

    • A shared repository with everything the team built, in git, with history
    • Each participant's sandbox for 30 days after, and an install script if they want it locally
    • A team playbook: prompts, guardrails, what to automate next
    awesome-ai-system-prompts

    The public reference collection of production system prompts, used as teaching material for prompt and agent design.

    See the build
    h2mai.com

    This site: a research build of a five-layer silver UI, WebGL cloth, shader type and a six-step intake, used as teaching material for the workshop.

    See case study

  4. 04 / 04 · Bespoke software

    Replace the per-seat licence with code you own

    Software cut to your process, and paid SaaS you rent by the seat rewritten into code you own.

    • Replace per-seat SaaS with owned code
    • Data platforms that pay per row
    • Crawling and extraction engines
    • Marketplaces and comparison sites
    Full page

    The right call when

    1. You pay per seat for a tool your team uses ten percent of.
    2. You buy data per licence that a crawler on your account could price per row.
    3. You need a marketplace or a comparison site where every number can be checked in public.

    You get

    • The paid tool taken apart: what you use, what you do not
    • The rebuild as code on your account, no per-seat bill
    • Crawling and extraction engines priced per row
    WhiteContext

    Data platforms that pay per row, not per seat: a Rust-to-WebAssembly extraction engine behind a vet marketplace.

    Read the case
    porownajdentyste.pl

    Dental price comparison built on an owned crawler instead of a paid data feed; every price checkable in public.

    Read the case
    Adcare

    Per-clinic CRM that replaced a per-seat subscription for an agency and its clinics.

    Read the case

How every engagement runs

Three steps. The six-step version, with the brief and the video request, is on the home page.

  1. 01

    Scoping call

    30 min, free

    You describe the situation; I ask three questions and propose a shape.

  2. 02

    Roadmap and architecture

    $1,250 deposit

    A written plan with a date. The deposit is credited to the first invoice; the documents are yours either way.

  3. 03

    Build, handoff, operate

    date on the roadmap

    One engineer builds every layer on your account, documents it, and can keep operating it.

Questions I get about this

Tap a question. The answer comes back like a text.

Scoping calls · most weeks

Maksym Huczyński, the engineer who built the systems on the Work page. I write the roadmap, the code and the reply to your brief. Agents do the volume; one person stays accountable for the whole system.

The roadmap gives you a fixed number before you commit to anything. Small changes start within days; a rebuild of a site, a CRM and the automations is weeks, with the date written down. The 30-minute call is free.

A written roadmap and an architecture for your system, delivered after the scoping call. If we build together, the full amount comes off the first invoice. If not, the documents are yours to take to anyone.

You do. The repository, the Cloudflare account, the domains and the data sit under your company from day one. I get access, not ownership.

The code, the accounts and the documentation are yours, and the system runs on Cloudflare without a server to babysit. Another engineer can pick it up from the docs I leave. That is also why every scheduled task leaves proof of its last run.

The system keeps running on Cloudflare with no servers to babysit. You can operate it yourself, keep me on for changes, or hand it to another engineer with the docs I leave behind.

Yes, before you send the brief if you need it. Send yours or use mine.

Both. Calls and documents in whichever you prefer. The code and the commits stay in English.

Something else? Ask it on the call

A 30-minute call, free

Thirty minutes, then a written plan

Leave an email. I reply within one working day with three questions and a proposed plan, not a deck.

Prefer to write the full brief first? Start there · maksym@h2mai.com · privacy notice