Case studies

What was broken, what was built, and what runs reliably now

Specific examples across contract automation, CRM rebuilds, creator discovery, reporting normalisation, and internal AI operating systems.

Lead build

Contract Automation

A closed-won deal triggered a fulfilment workflow: task context pulled from ClickUp, storage created in Drive, contract drafts generated from templates, fields updated, Slack notified, Sheets logged, and temporary state cleaned up.

  • Replaced manual post-deal fulfilment admin with a structured workflow.
  • Used templates, routers, datastores and explicit fallbacks instead of fragile one-step generation.
  • Designed it for handover: clear systems, alerts, storage conventions and an audit trail.
Read full case study →
Shipped, sanitised
  • 130Make modules (26 upgrades)
  • 62.9%faster contract cycle time
  • 2+hours saved per day
  • 6connected systems
Case study

Creator Discovery + Data Enrichment

Cross-platform creator discovery shipped: YouTube, Instagram, and TikTok

Built cross-platform creator discovery workflows for YouTube, Instagram and TikTok: scraper runs, profile enrichment, handle normalisation, deduplication against the influencer index, and quality filters before downstream use.

  • YTdiscovery + API enrichment
  • IGprofile enrichment
  • TTscraper workflow
  • Indexdedupe layer
  • Normalised creator handles and platform URLs before records entered the operating workflow.
  • Used scraper outputs, enrichment checks and index matching to reduce duplicate or low-quality leads.
  • Built reusable evidence for creator operations, data quality and cross-platform enrichment work.
Read full case study →
Case study

Autonomous Discovery & Evaluation Pipeline

Shipped, running daily

Built a multi-source autonomous pipeline that ingests records from 30+ web sources via zero-token scrapers, scores each against structured criteria with language model evaluation, deduplicates through a JSONL ledger, and generates privacy-safe documents with client names automatically aliased before any output is produced.

  • 22k+listing scans
  • 30+web sources
  • 300+records tracked
  • client-safeprivacy by design
  • Replaced manual multi-source scanning with automated zero-token ingestion using Playwright scrapers, the Greenhouse API, and WebSearch.
  • Built language model fit scoring against structured criteria to filter non-qualifying records before human review.
  • Enforced strict privacy gates by automatically aliasing client names as a media agency client prior to document generation.

Additional delivery work

More work across automation, data, and operations, including internal tooling and personal workflow systems.

FLOW
CRM Rebuild (TechCity)

Rebuilt CRM and sales infrastructure from scratch, delivering 35% faster sales cycles and 80% fewer lost leads across six portfolio businesses.

AI
SIRA Hybrid Retrieval & Memory Engine

Multi-stage hybrid retrieval with BM25 keyword matching, corpus enrichment, and sub-10ms context consolidation.

SYS
Personal OS Dashboard

Live task orchestration, forensic decision ledger, FTS5 memory search, and model quota management.

AI
Second Brain Knowledge Layer

277 notes embedded with hybrid search, MCP tools, and nightly consolidation.

FLOW
Make deployment governance

Safer versioning, deploy and handover for complex Make automations.

ENG
n8n infrastructure rebuild

17-phase Docker-containerised platform with AI-native connectivity.

OPS
ClickUp CRM from scratch

2 Spaces, 5 interconnected lists, cross-list relationships, per-list automations.

Want the relevant case-study angle for your workflow?

Send the handoff or tool stack. I will respond with the likely first fix and whether a sprint or audit is the right entry point.

Contact Simon