Notes on AI integration

Short, practical notes from projects I have built inside real businesses: what a forward deployed engineer does, how to hire a freelance AI engineer, how to wire an LLM into the software you already run, and where deterministic code beats a model.

1. 01
   What Is a Forward Deployed Engineer (and When a Company Needs One)Forward deployed engineer · 5 min

   A forward deployed engineer works inside the client's business, uses its real data and systems, and takes an AI solution to production. What the role is and when a company needs one.

2. 02
   Forward Deployed Engineer vs Solutions Engineer: What Is the DifferenceForward deployed engineer · 4 min

   Forward deployed engineer vs solutions engineer: who builds, who sells, who owns production. A side-by-side comparison and how to choose for an AI project.

3. 03
   How to Become a Forward Deployed Engineer: Skills and a Realistic PathForward deployed engineer · 5 min

   How to become a forward deployed engineer: the skills that matter, a realistic path from software engineering, and how to build proof of work without a job title.

4. 04
   AI Integrator vs AI Consultant: Who to Hire for WhatHiring · 4 min

   AI integrator vs AI consultant: one builds and wires AI into your systems, the other advises. A clear comparison, with a simple rule for choosing.

5. 05
   How to Hire a Freelance AI Engineer: Scope and Acceptance CriteriaHiring · 6 min

   How to hire a freelance AI engineer: define the task, ask for proof of shipped work, agree acceptance criteria and a price model for short projects.

6. 06
   How to Integrate an LLM Into Existing Business Software (CRM, POS, Messengers)LLM integration · 6 min

   How to integrate an LLM into existing business software such as a CRM, POS or messenger: where to put the model, what to keep deterministic, and how to test it.

7. 07
   Deterministic Forecasting vs an LLM: Why No Model Touches the NumbersForecasting · 5 min

   Why a forecasting system should use deterministic code and Monte Carlo simulation, not a language model, for the numbers. Reproducibility, back-testing and honest ranges.

8. 08
   Approval Rules as Data, Not Code: How Flexible Approval Workflows Are BuiltWorkflow systems · 5 min

   How to build a flexible approval workflow: keep the rules as data, route by category, project, author and amount, and protect it with an append-only audit log.

9. 09
   Public Records Aggregation: What to Build and What to BuyData aggregation · 5 min

   Public records aggregation from official sources: when to buy a data provider, when to build an aggregator, and how to design one so a failing source does not stop the rest.

10. 10
    AI Automation for a Small Business: Where to StartAutomation · 5 min

    AI automation for a small business: how to pick the first task, what to automate first, what to keep human, and how to get a result in a few weeks.