I install local language models inside your office and build the document automation on top — drafting, summarizing, note generation — with zero data egress and zero per-word API fees. Built for law firms, clinics, and anyone else who can't paste client files into a cloud chatbot.
30 minutes. I'll tell you honestly if local AI fits your workflow — and what it would cost.
Every cloud AI tool asks the same thing: send us your documents. For most regulated firms, that's the end of the conversation. It doesn't have to be — the models can come to you instead.
The ABA's guidance on generative AI requires lawyers to vet how AI tools handle confidential client information. With a local model the answer is: it runs on a machine the firm owns, on the firm's network, and no document ever leaves it. That's a sentence you can put in an engagement letter.
Cloud AI touching PHI means business associate agreements, security reviews, and ongoing vendor risk. A local deployment sidesteps the category entirely — no PHI is transmitted, so there's nothing new to disclose and nothing new to breach.
Per-token pricing scales with usage forever. A local stack costs what the hardware costs — often hardware you already own. Everything I ship runs at $0/month in API fees. The meter simply doesn't exist.
$ where-does-my-data-go --check cloud chatbot ......... 3rd-party servers "enterprise" AI tool .. 3rd-party servers + BAA local model (Ollama) .. your machine. full stop. $ monthly-api-bill --forecast cloud, per token ...... grows with every document local ................. $0.00 — no meter running $ what-it-actually-does ✓ dictation → structured clinical note ✓ intake email → client summary + draft reply ✓ document pile → extraction spreadsheet ✓ case notes → engagement letter draft ✓ feedback inbox → weekly themes digest all offline. all yours.
Exact price quoted after a scoping call — and it doesn't move after we shake hands. Code, prompts, and documentation are yours outright.
Find out exactly what to automate — with proof, not promises.
Your single most painful task, automated end-to-end.
A connected multi-step system for your core workflow.
Keep it fast, current, and quietly improving.
The guarantee: if the Audit doesn't identify at least one automation that credibly saves 5+ hours a week, the report will say so — and I'll refund half the audit fee. I'd rather keep the reputation than the invoice.
This isn't a pitch deck — it's a portfolio of working software, live on GitHub and Gumroad. Every tool runs on local models via Ollama with no API keys and no cloud. Your build is the next iteration of a stack I already run in production.
Turns dictation and rough notes into structured clinical notes — entirely on-device. PHI-safe by architecture, not by policy.
Healthcare wedgeMines hundreds of customer reviews for themes, complaints, and quotable lines. Same engine works on patient and client feedback.
Feedback analysisOne idea in, platform-ready drafts out. The drafting pipeline behind client updates, firm newsletters, and outreach.
Drafting pipelineScans and summarizes market and news data on a schedule. The watch-and-summarize pattern, productized.
Scheduled automationOne view of output, queues, and performance — the same dashboard layer that ships with every Pro build.
Reporting layerProduct info in, optimized Etsy/Shopify listings out, at scale. Bulk structured-document generation on a local model.
Bulk generationLogs decisions and runs AI-powered reviews against them. Structured record-keeping plus automated analysis.
Records + reviewThe eighth tool gets built for your workflow, on your hardware, to your spec.
Start the conversationI'm Hugo Kuznicki. I build AI automation on local models — the kind that runs on a machine you own instead of someone else's cloud. Seven shipped tools on that stack so far, all running at zero API cost, all built in public.
When you hire me, the person who scopes the work is the person who builds it, deploys it, and trains your staff on it. No account managers, no handoffs, no junior devs learning on your invoice. Fixed scope, fixed price, and a bias for shipping something your team is actually using within weeks — not a transformation roadmap for next year.
I document the whole journey in my newsletter and on X — you can watch how I work before you ever pay me a dollar.
30 minutes. You describe the workflow that eats your week — I'll tell you honestly whether local AI can fix it, what it would cost, and what I'd build first. If it's not a fit, I'll say so on the call.
kuznickicapital@gmail.com · replies within one business day