Launching MAP — Bangladesh's own maps platform

Bangladesh has 170 million people, 11 million daily ride-share trips, and one of the fastest-growing on-demand economies in South Asia. We've also been billing for that infrastructure in US dollars to a company that doesn't understand Gulshon means Gulshan, doesn't parse House 23, Road 11 as a holding number, and routes around mosques exactly never. That changes today.

MAP is a self-hostable, BDT-priced location platform built for how Bangladesh actually addresses. Geocoding that understands Banglish ("Gulshon" → Gulshan), Bangla script (ধানমন্ডি), and landmark phrases ("hospital near Banani"). Routing tuned for the vehicles Bangladeshis actually drive — including rickshaws. Webhooks for the geofence events that ride-share dispatch needs. APIs you can swap in without rewriting your app.

What's shipping today

• Geocoding: forward, reverse, batch — with the bdnlp normalisation pipeline on every query.
• Routing: directions (Bangla turn-by-turn), distance matrix, multi-stop optimisation, isochrones, GPS map-matching.
• Vector tiles: Bangladesh-only PMTiles bundle, MapLibre-compatible, Bangla labels by default.
• BD-specific endpoints: rickshaw-tuned routing, prayer-aware directions that detour around mosques during jamaat, and a 14-mosque registry for central Dhaka.
• Webhooks: geofence enter/exit, ETA thresholds. HMAC-signed delivery with exponential-backoff retries.
• SDKs: TypeScript, Python, Go, Swift, Kotlin, plus a Google Maps JS drop-in shim. Every endpoint covered, every SDK tested against the live gateway.

Why it's cheaper

Roughly 60–75% lower than Google Maps Platform at typical BD ride-share and delivery volumes. Two reasons: we're priced in BDT (no FX margin layered on top of a USD price), and our cost structure is BD-specific — single-region infrastructure in Singapore (closest Fly.io region), Bangladesh-only OSM extract instead of a global tile pyramid, and no enterprise sales motion to fund.

Why it's better for BD

It's not just translation. Google geocoding for "hospital near Banani" returns nothing useful. Banglish queries with phonetic spellings get the wrong locality. Holding numbers (House 23, Road 11) aren't parsed. We built bdnlp — a normalisation pipeline that runs on every query — to fix all of that.

What you can do right now

1. Search any BD address → (free, no signup)
2. Get a free API key → (25k requests/month, no card)
3. See the dispatch demo → (live ride-share matcher)
4. Read the 5-min quickstart →

What's next

Postal codes, Upazila/Union boundaries, Pathao / Foodpanda case studies, on-prem packaging for enterprise customers. See our roadmap.

Try it. Tell us what breaks. We'll fix it.