What this looks like in practice
One engagement where the work was advisory and organizational, and one where it was building and running the whole thing. Most engagements sit somewhere between.
AI-first transformation for an events, marketing and media organization
A business whose output is content and customer contact, carrying legacy infrastructure and a set of manual processes that did not scale with it. The brief began as AI integration and quickly became the broader question of what the technology organization needed to look like.
AI architecture and agentic automation
We architected and deployed an AI-first product roadmap, integrating LLM and agentic workflows across content publishing, automated customer support and lead qualification. The common thread was converting operational bottlenecks — work that waited on a person being available — into repeatable pipelines that run on their own.
Data governance and AI reliability
Putting AI into production surfaced the data problem underneath it. We established enterprise data governance, prompt evaluation standards and data quality protocols, so that outputs were reliable enough to publish, proprietary data stayed contained, and hallucinations were caught before they reached a customer rather than after.
Platform modernization and organizational redesign
Legacy infrastructure was moved into a scalable cloud environment, with modern cybersecurity standards established alongside it. That came with a restructuring of the IT organization itself — the new platform needed different roles and different responsibilities than the old one, and skipping that step is how modernizations get quietly reverted.
Capital efficiency
Third-party vendor stacks were reviewed and consolidated and internal workflows streamlined, driving operational cost savings and increasing the engineering output obtained per dollar spent.
Green Leaf Golf
A multi-tenant platform for running golf club competitions — live hole-by-hole scoring, automatic leaderboards, USGA handicap calculation and GHIN score posting. Designed, built and operated end to end, including the AWS infrastructure and the iOS and Android apps. It is where the hands-on half of the practice is kept sharp.
- Role
- Sole engineer — product, build and operations
- Backend
- Django 5, Python, PostgreSQL 15
- Infrastructure
- AWS ECS Fargate, RDS, load balancing, S3, CloudWatch — defined in Terraform
- Mobile
- Native iOS and Android, submitted to both stores
- Live at
- nbs.golf
Scoring that survives the course
Golf courses have dead spots, and a scoring screen that needs a round trip for every hole stalls exactly where the signal is worst. The scoring flow holds the whole round on the phone: it downloads once, renders every hole locally, and queues score writes until there is signal. Writes carry the time they were entered rather than the time they arrive, so a phone flushing twenty minutes of backlog cannot silently undo a correction someone already made.
Operations
Infrastructure is defined in Terraform and deploys through a script that verifies the built image before it touches the running service and reports exactly which revision to roll back to. Alerting covers the four conditions that mean something is genuinely broken. CPU, memory and latency alarms are deliberately absent — they fire during a busy event and clear on their own, which is how alerting stops being read.
Built to be handed over
Every non-obvious decision is written down beside the code — not what it does, which is readable, but why it is that way and what breaks if it changes. Over two hundred tests cover the paths that have caused real incidents.