The work I take on: an assistant that reads your files, AI wired into the tools you run, the recurring job that runs itself, a portal your team signs into, your data in one table you trust, and every enquiry answered.

View work hire@scrocle.cloud

Work

Selected work at full size. Every entry opens as a full case study.

All 14 case studies

Shipped and running systems first. Laboratory builds and work in progress are labelled.

01In production

AI

Autonomous AI agent operations

Scheduled AI agents on Hermes, MCP, and LiteLLM that call your databases and APIs, fail over across models, and alert your team when a run needs a human call.

  • Hermes
  • MCP
  • LiteLLM
  • cron
02Shipped

Automation

Process automation suite

Fifteen n8n workflows connecting CRM, email, and accounting over REST and webhooks, with a logged error path on every flow so failed records can retry.

  • n8n
  • REST APIs
  • Webhooks
03Shipped

Data

E-commerce data pipeline

A daily job that collects competitor prices from storefronts and writes a table. Retries when a site changes. Nobody opens a browser at 9am to do it.

  • Python
  • BeautifulSoup
  • Selenium
04Shipped

Data

Sales data analysis

Seven months of sales records turned into a model and a dashboard. The client still opens it each quarter.

  • Power BI
  • SQL
  • Excel
05Shipped

Data

Technical SEO audit tool

Eleven on-page and crawl checks, then a ranked list of fixes. Built to run again after a template change, without a monthly SEO product.

  • Python
  • BeautifulSoup
  • JSON
06Shipped

Automation

Content production pipeline

A queue for scheduled publishing across a few brand voices, with a check before anything is marked ready. The client still owns the writing.

  • Python
  • n8n
  • SEO
07In progress

Desktop

PixPolorer

A Windows program for a large photo and video library that already lives on the computer, including files that should never go to the cloud.

  • Qt 6
  • C++
  • Rust
  • SQLite
08In progress

Android

Hermes Android

A phone app for talking to an AI agent you run yourself. Built around a locked session with a real terminal.

  • Kotlin
  • Jetpack Compose
  • WebSocket
09In progress

Android

PwnLink

A phone app for running a Raspberry Pi when you are away from a desk. Bluetooth, location, a connection that survives the screen turning off.

  • Kotlin
  • Bluetooth
  • GPS
10Laboratory

Data

Ingestion

A pattern for fetching pages, checking them, dropping duplicates, and parking bad records so they never mix with good ones.

  • Python
  • Camoufox
  • SQLite
  • Docker
11Laboratory

AI

Telegram agent walkthrough

An interactive walkthrough of a production Telegram agent architecture covering session memory, intent routing, specialist handoff, and model failover. Click a preset scenario to inspect each stage.

  • Python
  • FastAPI
  • MCP
12Laboratory

Automation

Jobs Engine

You type a title, skills, and location. The page ranks a frozen set of public remote jobs, shows why each one scored, and downloads the same rows as Excel. Matching is a keyword rubric, so every point is traceable. You can click it.

  • Python
  • FastAPI

Skills

  • AI chatbots & assistants

    Assistants that answer from your own documents, qualify what comes in, and hand over when judgement is needed.

    • LLM & RAG pipelines
    • Chatbot & assistant development
    • Natural language interfaces
    • Conversation memory & hand-off
    Show 3 moreShow less
    • Prompt engineering
    • Embeddings & semantic search
    • FastAPI
  • AI integration

    Your existing tools kept in step with each other, without replacing the ones that already work.

    • API integration
    • Model routing & failover
    • MCP (Model Context Protocol)
    • Tool calling
    Show 3 moreShow less
    • Webhooks & event automation
    • OpenAI-compatible APIs
    • LLM operations
  • Business automation

    Scheduled work that runs without supervision and reports what it could not finish.

    • Workflow automation
    • Scheduled automation
    • Alerting & notifications
    • Retry & error handling
    Show 3 moreShow less
    • API integrations
    • Webhooks
    • n8n
  • Web products

    Web applications with real accounts, roles, and rules that hold under load and misuse.

    • Full-stack web development
    • Next.js / TypeScript
    • React
    • Auth & user roles
    Show 3 moreShow less
    • Databases & SQL
    • Web APIs
    • Server-side security
  • Desktop & Android

    Desktop and Android software that keeps data on the machine and works without a network.

    • Native desktop apps (Qt / C++)
    • Android development (Kotlin / Compose)
    • Rust (systems programming)
    • SQLite & local databases
    Show 2 moreShow less
    • Data encryption & secure storage
    • Desktop & mobile engineering
  • Data systems

    Data from the web or your own systems, collected on a schedule and turned into something trustworthy.

    • Python / SQL
    • Data analysis & reporting
    • Data pipelines & automation
    • Web scraping & data extraction
    Show 3 moreShow less
    • Power BI
    • Browser automation (Playwright)
    • Docker & deployment

How a job starts

  1. 1
  2. 2
  3. 3
  4. 4
  1. 1

    Your brief on /hire

    Write down what is failing, the result you want, and your deadline. Add the systems you already run and pick a scope stage rather than a dollar figure.

  2. 2

    A fixed quote or a plain no

    I read your brief before we schedule any call. If the job does not suit my practice, I say no up front. If it fits, you receive a written scope and a fixed price.

  3. 3

    Short passes and a decision record

    I ship working software in short passes and write down every technical call and its cost. Anything discovered mid-build is priced before I touch it, so you decide first. Every case study on /work shows this decision log in full.

  4. 4

    Source, runbook, and team walkthrough

    You get the repository, deploy steps, credential map, and a walkthrough of the parts your people will touch. Your team runs and modifies the software on your own infrastructure without keeping me on retainer.

You keep the source, the documentation, and the runtime.