# BorikDev

> Nathan Borik builds custom AI systems and integrates LLMs (mainly Claude) into solo operators, small businesses and startups. Work starts with a free audit call, followed by a written proposal with one fixed price sized to the business, then build and hand-over. Remote, worldwide; based in Israel.

Who it is for, in priority order:

- Solo operators and micro businesses: consultants, freelancers, clinics, studios, salons, small shops.
- Small businesses: offices and service businesses with email, documents and customer messages to handle.
- Startups (seed to Series B): teams that want Claude connected to their whole stack via Model Context Protocol (MCP).

Typical builds:

- An assistant that knows a business's services and policies and drafts replies to email, WhatsApp or website chat for the owner to approve.
- Booking reminders, flagged cancellations and a daily summary for clinics, studios and salons.
- Customer-question answers drafted from shop policies and order data, plus a weekly Shopify sales report.
- Invoices and documents read from email, filed into the right Google Drive folder, key numbers sent to the bookkeeper.
- For startups: Claude connected to GitHub, Jira, Confluence, Slack, Postgres, MongoDB, Mixpanel, AWS, Stripe, HubSpot, Intercom, Figma and more; database questions answered in plain language; scheduled reports and briefs.

Pricing: no public price list. Every project is quoted as one fixed price after a free audit call. No hourly billing.

Contact: admin@borik-dev.com · Book a free audit call: https://cal.com/nathan-borik-hdtplm/l60m · LinkedIn: https://www.linkedin.com/in/nathan-borik-a3695463/

## Main pages

- [Home: custom AI systems for small businesses and startups](https://borik-dev.com/): services, examples, how it works, pricing model, FAQ
- [Technical interview and recruitment services](https://borik-dev.com/hiring): hiring help for technical roles

## Writing

- [The 4-layer prompt cache fix that cut LLM costs by 82%](https://borik-dev.com/blog/llm-prompt-cache-cost-reduction): diagnosing and rebuilding prompt-cache architecture, with production numbers
- [Why I send the full conversation history on every Claude turn](https://borik-dev.com/blog/claude-full-conversation-history): the mental model that makes prompt caching work
- [Multi-layer LLM memory architecture for conversational AI](https://borik-dev.com/blog/llm-memory-architecture): history window, personality profile, decisions window and vector retrieval
- [Making your entire codebase readable to AI](https://borik-dev.com/blog/ai-native-code-readability): why AI-native code readability matters more than prompt engineering

## Optional

- [Fractional CTO for non-technical founders](https://borik-dev.com/fractional-cto-non-technical-founders)
- [Technical cofounder alternative](https://borik-dev.com/technical-cofounder-alternative)
- [Startup technical validation](https://borik-dev.com/startup-technical-validation)
- [Architecture review before building](https://borik-dev.com/architecture-review-before-building)