- Problem
- Operators need to spot degraded positioning signals before they become expensive surprises.
- What I built
- A monitoring workflow for ingesting, scoring, and surfacing GNSS behaviour in near real time.
- Stack
- Python, time-series processing, alert logic, dashboards, ops notes.
- Impact
- Turns specialist signal behaviour into visible operational signals.
- Why it matters
- Good alerts let teams act early instead of debugging from wreckage later.
Data / Python / GNSS / Forward-deployed engineering
Luke Howlett — Python, Data, GNSS & Useful Engineering
Hands-on data and research engineer building tools, dashboards, infrastructure, and analysis systems that turn messy technical work into something people can actually use.
Interactive skills map
Click a lane to see where the evidence shows up.
streaming data pipelines
GNSS quality monitoring
analysis automation
Docker and Nginx deployments
reproducible VPS ops
interactive dashboards
semantic search products
stakeholder-ready demos
technical explanation
Featured proof projects
Evidence of the kind of systems I like building.
- Problem
- Correction streams can drift, disagree, or hide subtle biases that affect downstream positioning.
- What I built
- Analysis tooling to compare stream behaviour, expose bias patterns, and make review repeatable.
- Stack
- Python, pandas-style analysis, plotting, GNSS domain checks.
- Impact
- Gives researchers and engineers a faster route from raw corrections to evidence.
- Why it matters
- Bias analysis is where "it looks fine" becomes measurable confidence.
- Problem
- Personal media and memories become hard to search once they leave neat folders and labels.
- What I built
- A semantic search direction using image/text embeddings to make photos and moments findable.
- Stack
- Python, embeddings, search indexes, lightweight product UI.
- Impact
- Moves a pile of media toward a useful, searchable personal knowledge system.
- Why it matters
- Search is most valuable when people can use natural language, not perfect filing habits.
repobuildreleasenginx
- Problem
- A live VPS had useful services, but too much state lived only in the server's memory.
- What I built
- A repo-owned deployment model with inventory, Nginx templates, safe releases, backups, and runbooks.
- Stack
- Make, Bash, Nginx, Docker Compose, Certbot, WireGuard docs.
- Impact
- Turns a hand-maintained VPS into infrastructure that can be audited and rebuilt.
- Why it matters
- Small systems deserve grown-up operational habits too.
- Problem
- Planning a DJ set is part music taste, part sequencing, and part not losing your flow.
- What I built
- A web app and deployment wrapper for shaping playlists and making the tool available on the VPS.
- Stack
- React/Vite-style frontend, Python API, Docker Compose, Nginx routing.
- Impact
- Shows product instinct: build the workflow, ship it, and make it reachable.
- Why it matters
- Useful tools often start as "I wish this existed" and become proof of delivery.
How I work
Comfortable where the problem is still a bit untidy.
Curious. I like understanding the domain, not just wiring code together.
Pragmatic. I care about clean systems, but I care more about getting useful things into people's hands.
Good with messy systems. Legacy VPS, odd data, unclear ownership, half-documented workflows: fine, let's map it.
User-facing. I am comfortable talking to users and stakeholders, then turning that into working software.
Make it real. I prefer demos, deployments, dashboards, and evidence over abstract promises.
Contact
Want the human version?
Here are the proper links. Replace the placeholders when ready.