'Near the big masjid, blue gate' — unstructured addresses break delivery, emergency response, and credit verification
I have geocoded flood zones, election constituencies, and delivery territories, and everywhere the same foundation is missing: most Pakistani addresses are directions, not addresses. 'House 12/B, street of the tandoor, near Faizan Masjid, Orangi' — readable by a human neighbor, useless to a routing engine, ambiguous to an ambulance at 3 AM. The costs surface everywhere at once: couriers average multiple calls per delivery and fail 10-15% outright; e-commerce COD verification cannot confirm an address exists; banks reject loan applications because the address cannot be verified; rescue services lose minutes they do not have. Global solutions (plus codes, what3words) exist but have no local adoption loop. The wedge is B2B: a Pakistan-tuned address resolution API trained on how people actually write addresses — landmark-relative, mixed Urdu-English, phonetic spellings — sold to the couriers and banks already bleeding money on this, while quietly building the national address graph underneath. Every delivery confirms and enriches the data. I have built the prototype geocoder; it resolves 70% of a courier's problem addresses. This platform is where I am looking for the team to take it from model to company.
A high-conviction problem with strong founder-market fit signals. The combination of severe price asymmetry, accessible demographics, and existing infrastructure makes this buildable within 9 months by a small team.
Solutions · 4
AddressLink: landmark-relative address resolution API trained on Pakistani address language, sold courier-first
Formalizing the plan behind my prototype. The engine: an NLP model trained on millions of real delivery addresses (how Pakistanis actually write — 'makan 12/B gali nanbai wali, Faizan masjid k saath, Orangi 11½'), resolving to rooftop-level coordinates with a confidence score, improving from every confirmed delivery. Product one is a courier API: address-to-pin resolution plus failed-address rescue queues, priced per successful resolution — the couriers' own failure costs make the ROI case in one meeting. Product two, six months later: the address registry itself — households claim and verify their AddressLink code, which then works across every integrated courier, bank KYC, and ambulance dispatch. The graph compounds; the moat is the confirmed-delivery feedback loop no global player has. Build Space open: need a senior backend engineer, a data annotation lead, and a courier-sales operator.
Rider-sourced ground truth: pay per confirmed rooftop pin during normal delivery runs
The training data the resolution engine needs is collectable by the people who solve addresses manually every day: riders. Add a confirm-pin step to delivery apps — rider drops the verified rooftop pin at handoff, earns PKR 5-10 per confirmed new address, cross-validated by GPS trace and occasional photo audit. A single courier's fleet confirms tens of thousands of addresses monthly as exhaust from work already done. Our riders would treat it as a bonus program; I can pilot the workflow inside my company's operation the week the API is ready to receive.
Emergency services integration from day one: the address graph's highest-stakes user should shape the schema
Design the ambulance case in from the start, not as CSR garnish later: a 1122 dispatcher needs resolution from a panicked caller's spoken description, offline fallback maps for signal-dead katchi abadis, and pre-verified codes for the settlements where every second of confusion is lethal. The flood-response mapping I worked showed how quickly informal settlements become invisible in crises. Give emergency services free integration and priority schema input; the legitimacy this buys will open the government datasets the commercial product will eventually want.
Diaspora KYC angle: verified home addresses unlock remote account opening and property services
From the overseas files I manage: half of diaspora banking and property friction reduces to 'prove the address'. A verified address registry becomes the KYC rail for remote bank account opening, utility transfers, and the property-management services my relatives keep needing — the bank accepts the AddressLink verification instead of demanding a cousin's utility bill and a prayer. Banks pay real money for KYC cost reduction. Worth a dedicated business line once the courier wedge proves the graph.
Discussion
Dispatch supervisor: my riders' voice notes are a folk address-resolution system — 'the gali with the electric pole that leans'. Forty of them would map faster than any survey team. Use us.
Huzaifa — rider folk knowledge is literally the training corpus. The confirm-pin program in solutions pays your forty riders for what they already know. Dispatch-level partnership conversation, please.
Bank operations: address verification failures kill loan files daily — applicants who demonstrably exist at addresses we cannot formally verify. The KYC vertical alone justifies this company.
Flood response 2022: we delivered aid by asking mosques to announce names because addresses failed entirely. Emergency addressing is not an edge case — it is the case that matters most.
Seller: my order form now demands two landmarks and a WhatsApp location pin, and STILL 10% fail. I have become an amateur cartographer to sell phone covers. Godspeed to this Build Space.
Overseas angle: diaspora relatives cannot even describe their own property locations formally — 'the house near Chaudhry sahib's dera'. Title records and addresses fail together. Two platform ventures, one root cause.
Karachi courier here. The 11-minute failed-address average quoted matches my daily life exactly. Some days I am 40% navigator, 60% detective.
Marketplace data: address-ambiguity cancellations cluster in exactly the high-growth low-income zones where the next 50M customers live. Solving addresses IS market expansion, not operations hygiene.
Dadu: flood aid in 2022 missed our lane entirely because it exists on no map — we watched trucks serve the addressable and leave. Addressing is not logistics here; it is whether the state can find you when the water rises. Priority-list us.
The NLP challenge here is genuinely interesting — code-switched, landmark-relative, phonetically spelled. If the Build Space wants a junior pair of hands on the normalizer, my weekends are available.