Available for new opportunities


|

I eliminate manual work. Python automation, AI systems, and full-stack tooling that gives time back to the people who need it most.

0Production systems
0Hours R&D
0Automation accuracy
24/7AI uptime
Scroll

01 / 04

3 Months of Manual Work.
Automated in 5 Minutes.

Python parallelisation across 8 workers, extracting 2,547 assets from a legacy system with 99.8% accuracy.

PythonParallelizationAutomationData Compliance
0
Assets
5 min
Runtime
8
Workers
0
Accuracy
View Case Study →

02 / 04

End-to-End Reporting. Zero Manual Steps.

API integration with legacy database auth, recursive data extraction, automated report generation for 2,340+ records.

PythonAPI IntegrationData CurationAutomation
2d→4m
Time Saved
0
Records
0
Automated
0
Manual Steps
View Case Study →

03 / 04

From Form to Branded PDF
in Under 5 Minutes.

FastAPI backend, React frontend, Playwright-powered PDF generation. Production-deployed. A 60-minute manual task reduced to a button click.

PythonFastAPIReactVitePlaywright
60m
Before
<5m
After
0
Manual Steps
1
Click
View Case Study →

04 / 04

24/7 Autonomous
AI Orchestration.

Custom platform routing tasks between 5 specialist AI agents across 19+ integrations. 250+ hours of R&D. Runs while you sleep.

Context OptimisationRoutingWorkflow DesignPythonWebSocket
0
Hours R&D
19+
Integrations
24/7
Uptime
5
Agents
View Case Study →

Contact

Get in touch.

Open to automation projects, full-stack builds, and AI integration work. Based in Bristol - available remotely.

Location
Downend, Bristol BS16
Web
josho.pro
Send me an email

What I build

Python Automation
Turning days of manual work into minutes. Data migration, extraction, transformation pipelines.
AI Orchestration
Multi-agent systems that plan, delegate, and execute complex tasks autonomously. 24/7, no hand-holding.
Full-stack Tooling
Production-deployed web applications. FastAPI backends, React frontends, real workflows.
Data Pipelines
End-to-end data collection, curation, and reporting. From legacy APIs to clean, usable outputs.

01 / 04  ·  Data Automation

3 Months of Manual Work.
Automated in 5 Minutes.

A legacy system held 2,547 critical assets locked behind a slow manual export process. This script parallelised the extraction across 8 workers, completed the job in under 5 minutes, and achieved 99.8% accuracy.

0
Assets Extracted
4m 53s
Total Runtime
8
Parallel Workers
0
Accuracy
Scroll to explore

How it works

01

Connect to legacy system

Script authenticates with the legacy database using credentials passed via secure environment variables. No hardcoded secrets.

02

Enumerate all 2,547 assets

Recursive query walks the full asset tree, building a manifest of every item to be extracted before any transfer begins.

03

Dispatch to 8 parallel workers

Python's concurrent.futures splits the manifest across 8 threads. Each worker extracts independently, updating a shared progress counter.

04

Validate and write output

Every asset is checksummed on arrival. Failures are flagged for retry. Final report confirms extraction count and accuracy rate.

Live demo

extract.py -- running
£ python extract.py --workers=8 --output=./export
✔ Worker pool ready (8 threads)
✔ Legacy system: connected
✔ Asset manifest: 2,547 items
Extracting... 0%  0/2547

What this would cost otherwise

ApproachEstimated Cost
Manual in-house (90 working days, UK mid Python dev avg £52k/yr)~£17,500 in staff time
Data migration agency (small-to-mid scope)£8,000–£25,000 project fee
Off-the-shelf migration tooling (annual licence + setup)£2,000–£8,000/yr
This solutionBuilt once. Runs in 5 min. Reusable.

Stack

Lang

Python 3.11

concurrent.futures for the worker pool, custom retry logic for failed checksums.

I/O

Direct DB + filesystem

Reads from legacy system over authenticated connection. Writes to structured output directory with manifest JSON.

QA

Checksum validation

Every asset verified on write. Any mismatch flagged and retried automatically before the run ends.

02 / 04  ·  Reporting Automation

End-to-End Reporting. Zero Manual Steps.

A reporting process that consumed two full working days per cycle. Now it runs in 4 minutes. 2,340+ records extracted, curated, and formatted automatically every time.

2 days → 4 min
Time Per Cycle
0
Records Processed
0
Automated
0
Manual Steps
Scroll to explore

How it works

Job reports, pulls all completed services for any and all specific sites, takes all notes, service status and automatically creates a .xlsx file to a template that's ready to ship.

Live pipeline

report_pipeline.py
Auth
Extract
Curate
Report
Waiting for trigger...

What this would cost otherwise

ApproachEstimated Cost
Staff time: 2 days/cycle × 26 cycles/yr (UK avg office worker £32k/yr)~£6,700/yr in labour
Outsource to reporting/BI agency£500–£2,000 per report run
Tableau or Power BI (Professional + setup consultant)£4,200–£14,000/yr + setup
This solutionZero ongoing cost. Runs on demand.

Stack

Auth

Custom session handler

Wraps the legacy API's non-standard token flow with automatic refresh and retry on 401.

Data

Recursive pagination

Handles variable page sizes, cursor-based and offset-based pagination in the same codebase.

Out

Structured output

Normalised schema, consistent formatting, ready for downstream consumption or direct delivery.

03 / 04  ·  Full-Stack Tool

From Form to Branded PDF
in Under 5 Minutes.

A production-deployed tool that takes client and report data via a web form and generates a fully branded, multi-page PDF in seconds. FastAPI backend, React frontend, Playwright-powered PDF engine.

60m → <5m
Generation Time
Live
Production Deployed
0
Manual Steps
1
Click to Generate
Scroll to explore

How it works

01

React frontend (Vite)

User fills a clean web form. Client name, report title, date, and executive summary bullets. No formatting required.

02

FastAPI receives the POST

FastAPI endpoint receives the form data, merges it with the JSON data template, renders the document via headless browser.

03

Playwright-powered PDF

Headless Chromium renders the HTML with branded CSS. Supports modern layouts, page numbers, and complex charts.

04

PDF returned instantly

The binary PDF is streamed back to the browser as a file download. No files saved server-side. Generation time: under 5 seconds.

Try it live

input form
Client Name
Report Title
Date
Executive Bullets (one per line)
↟ Updates the preview live — no button press needed
output_report_04.pdf — 100%

Strategic Growth Review 2026

PREPARED FOR: Meridian Partners

2026-02-27

Executive Summary
  • Projected revenue uplift of 23% within 12 months
  • Priority expansion: Nordics and DACH region
  • Headcount growth of 18 planned for H2
Quarterly Progress
Q4
CONFIDENTIAL PAGE 1 OF 4

What this would cost otherwise

ApproachEstimated Cost
Manual document creation (1hr/report, 50 reports/yr)£2,500–£5,000/yr staff cost
PandaDoc / DocuSign enterprise plan£3,000–£8,000/yr
Custom dev agency (FastAPI + React + PDF)£8,000–£25,000 build
This tool (production-deployed, reusable, fully owned)Built. Running. Yours.

Stack

Back

FastAPI & Playwright

High-speed async API receives form data and drives headless Chromium to render and capture the branded PDF.

Front

React (Vite)

Lightweight, high-performance UI for data entry with live PDF preview updating in real time.

Logic

Jinja2 Templating

Dynamic HTML generation with multi-page support, auto-calculated table rows, and brand colour injection.

04 / 04  ·  AI Infrastructure

24/7 Autonomous
AI Orchestration.

250+ hours of R&D. A custom platform routing tasks between 5 specialist AI agents across 19+ integrations. It thinks, delegates, builds, and delivers - while you sleep.

0
Hours R&D
19+
Integrations
24/7
Uptime
5
Specialist Agents
Scroll to explore

You describe a task. The AI figures out who's best at it, delegates it, builds it, and sends it back. No setup. No instructions. It just works.

You
AI AssistantOrchestrator
Research AI
Design AI
Deliveredvia Telegram
01

Your Request

"I need to do a presentation for work... can you understand my files in /Documents/Task and build me one?"

02

Triaged by AI Assistant

The orchestrator analyses the request, reads the 3 related files, and writes detailed briefs for the specialist agents.

03

Parallel Execution

Research AI extracts 7 key insights and drafts a slide structure. Design AI builds a custom PPTX applying your brand colours (#1A3C6E).

04

Automated Delivery

The orchestrator synthesises the work and delivers the final presentation directly to your phone. Ready for the meeting.

AI Assistant
Orchestrator
Triages tasks, picks the right specialist, writes briefs, synthesises results.
Research AI
Research
Deep research, file analysis, competitive intelligence, structured reports.
Build AI
Engineering
Writes, debugs, deploys code. Full-stack builds, scripts, infrastructure.
Design AI
Design
UI/UX specialist. Self-contained HTML prototypes, design systems, visual assets.
Automation AI
Automation
Cron jobs, data pipelines, integration glue across 19+ connected services.

What this would cost to build or buy

ApproachEstimated Cost
Reactive AI agent (chatbot, off-the-shelf, rule-based)£16,000-£28,000
Intermediate agent (multi-step workflows, API integrations)£32,000-£56,000
Advanced autonomous agent (planning, tool orchestration)£64,000-£96,000
Enterprise multi-agent system (swarms, legacy integration)£80,000-£160,000+
Hire equivalent team (5 specialists, UK market salaries)£250,000+/yr ongoing
This system - custom-built, 250+ hrs R&D, production-liveBuilt. Running. 24/7.