A small senior team building geospatial software that'ships and stays shipped.
Antef started in 2016 as a small group of engineers building survey and mapping software for the public sector in India. A decade in, we work across insurance, utilities, research and consulting — and we're expanding the practice to serve more buyers in the US, UK and EU.
The principles, plainly.
01
Outcomes, then artefacts.
Every engagement is scoped against a measurable change. The dashboard, the model, the report — those are the means.
02
Senior engineers, named.
Whoever pitches the work is on the work. No anonymous resourcing, no body-shop rotation, no juniors learning on your dollar.
03
Open standards by default.
PostGIS, MapLibre, GeoServer, OGC standards. We use commercial layers when they earn their keep, never out of habit.
04
Honesty about scope.
If your problem isn't ours to solve, we'll say so and point you somewhere better. The roster of firms we trust is part of the value.
Productised. Priced. Done.
We’ve moved away from time-and-materials engagements wherever we can. Most of our work now ships as fixed-scope packages with a price you see before we start: discovery, pilot, full delivery. It’s harder for us — and better for every buyer.
We also don’t scale by hiring junior contractors and charging the difference. The team is small, senior and named on engagements. When we don’t have the right person in-house, we’ll tell you and recommend someone we trust.
Honest about what AI does and doesn’t do.
Most “AI” in geospatial is rules and human review with a marketing layer on top. We’re trying to be the boring exception.
01
Where ML earns its keep.
Building-footprint extraction, change detection on imagery, vegetation encroachment around assets, roof-condition classification. These are problems where models trained on labelled data outperform rule-based heuristics. We use ML there.
02
Where we don’t pretend.
We don’t wrap a generic LLM in a map UI and call it geospatial intelligence. We don’t claim “AI” on workflows that are actually rules, regexes and human review. If a customer wants a chatbot, we’ll build them a chatbot — and tell them it’s a chatbot.
03
Honest GeoAI.
Where we are using LLMs — see GeoChat on the products roadmap — the model writes the structured spatial query and the database answers it. The analyst sees both. There’s no hidden geometry hallucination, no “trust me” layer between the question and the data.
Open by default. Commercial when it earns it.
We contribute upstream where it has value and reuse what already works. The customer’s stack outlives our engagement.
- PostGISStorageSpatial database underneath nearly everything we ship.
- GeoServerServingOGC-compliant tile and feature serving for analyst-facing systems.
- pg_tileservServingLightweight vector-tile serving direct from PostGIS.
- MapLibre GLClientDefault web map client. Open-source, vector-tile native.
- deck.glClientWebGL data visualisation when the layer count outgrows MapLibre.
- QGISDesktopAnalyst desktop tool. Where most field-side work actually happens.
- OGC standardsStandards bodyWMS, WMTS, WFS, OGC API Features. The interop layer that keeps stacks portable.
Offices, time zones, footprints.
01
Fazilka
India
Engineering
Want the longer version
Coffee, video call, or a written proposal.
We're easier to evaluate in conversation than from a page. Pick whichever costs you less time.