Custom systems & graphs

Work with me

I'm Maxim, an independent systems researcher. I build and run production data systems end to end — backend, data pipelines, and frontend — and I keep them online long enough to be worth trusting: five of them run right now off a single box in the EU, all open to inspect. The work on this site is not a portfolio of mockups; it is live software with real data flowing through it. If you have data and want a system built around it, that record is the pitch.

OFFERS

3 areas · fixed scope or weekly

Custom intelligence dashboards

You have data — an API, a database, a stream, a pile of CSVs — and you need a system that turns it into something people can actually read and act on. I take it from source to a live dashboard: a backend that fetches and stores the data on a schedule, a model or scoring layer if you need one, and a frontend that loads fast and says something. No template dashboards, no chart dumps.

What's included

Data ingestion and scheduled refresh (FastAPI + Postgres/PostGIS, or SQLite for lighter loads)
A scoring / modelling layer where the numbers need interpreting, not just plotting
A frontend built to fit the data — static HTML for a fixed cockpit, or Next.js when it needs to be interactive
Deployment to your infrastructure with health checks and a documented refresh cycle
Handover: source, schema, and a short runbook so your team can keep it running

Proof
GeoSignal Finance → · Oil Stress Monitor →

Engagement: fixed scope for a defined dashboard; weekly for ongoing model and data work.

OSINT / geospatial data pipelines

Open-source signals are noisy, rate-limited, and formatted for machines, not maps. I build the pipeline that pulls them, cleans them, stores them geospatially, and serves them to a map fast enough to be read in real time. This is the harder half of any OSINT system — the map is easy once the data is honest.

What's included

Integration of open feeds: ADS-B aircraft, AIS vessels, NASA FIRMS fires, GDELT/UCDP conflict events, satellite TLE propagation
Provider fallback and caching so a single source outage doesn't take the system down
PostGIS storage with a tiered cache for expensive or rate-limited APIs
Vector tiles (PMTiles) for static infrastructure layers served over HTTP range
A MapLibre / deck.gl frontend layer, or a clean API if you already have a map

Proof
OSINT Globe →

Engagement: fixed scope per data source or layer; weekly when you're standing up a multi-source system.

Knowledge-graph engineering

Most "knowledge graphs" are a spreadsheet with arrows. I build the real thing: an ontology that holds up, an evidence-based taxonomy where every claim carries its provenance, and — where it fits — publication into a decentralised knowledge protocol so the graph is queryable and portable, not locked in one database.

What's included

Ontology and schema design: entities, relations, and the property model that keeps them consistent
Evidence-first classification — every fact linked to a source, with a transparent, documented scoring heuristic (not a black-box score)
Deduplication and entity resolution so the same thing isn't in the graph three times
Publication to Geo Protocol (geobrowser.io) with idempotent, re-runnable batch scripts and audit snapshots
A GraphQL-backed view or static dashboard over the resulting graph

Proof
AI Atlas → · Agent Taxonomy Graph →

Engagement: fixed scope for an ontology + initial publish; weekly for ongoing curation and graph growth.

HOW I WORK

One pair of hands

You work directly with the person writing the code. No account layer, no handoff to a junior, no lost context.

Direct communication

Async by default, a call when it's faster. I say what's hard and what won't work before it costs you time.

Weekly demos

You see a running thing every week — not a status slide. Progress is measured by what's online.

Your code stays yours

Full source, schema, and a runbook at handover. Nothing you depend on lives only on my machine.

Have data that needs a system?

If you have a dataset, a signal problem, or a graph that needs building — tell me what the data is and what you need to see; I'll tell you honestly whether I'm the right person to build it.

EMAIL → @maximl ON X →