Case study · Founder & Architect
Smart technology to help reunite lost pets with their families.
A production platform that brings lost, found, and adoptable pets from shelters, SPCAs, and the public into one searchable map — and uses AI photo matching to connect lost pets with the animals that might be them.
When a pet goes missing, the first hours matter most — yet the information owners need is scattered across dozens of shelter websites, municipal pounds, and fast-moving Facebook groups. Found animals sit in a shelter listing while their owners search somewhere else.
Pawlert brings those sources together in one place, keeps them in sync automatically, and does the comparison work for the owner: every lost report is matched against every nearby found or sheltered animal, by photo as well as by description.
I founded Pawlert and own every layer of it — product direction, architecture, implementation, and operations. The key decisions:
1. Strategy
A lost-pet site is only useful if the pets are already there. Instead of waiting for users to post, Pawlert aggregates the public listings of shelters and SPCAs, so every search returns real animals from day one — and every new report is matched against them immediately.
2. Architecture
Reports and partner imports flow through RabbitMQ into a single pipeline: insert, geocode, compute a photo embedding, then match. Imports are keyed on source and external ID, so a nightly re-sync never duplicates an animal, and a failing third-party call dead-letters instead of breaking a user's report.
3. Engineering
Matching combines CLIP image embeddings with distance, date, colour, and gender into a weighted score, with exact microchip matches short-circuiting to 100%. Pairs missing a photo embedding score neutrally rather than disappearing, and are rescored automatically once the embedding arrives.
4. Execution
Infrastructure as code with Terraform, blue/green deployments, backward-compatible database migrations for zero-downtime releases, nightly backups, monitoring, image moderation, and a Cloudflare security layer — run by one person, built to the standard of a team.
Owners can search with a single photo. Pawlert computes an image embedding and ranks nearby animals by visual similarity, combined with distance and the other listing signals.
The same engine runs in the background on every new report, so a sheltered cat imported overnight is surfaced to the family looking for it — without either side having to find the other first.
Each source has its own integration — scheduled, monitored from an admin console, normalized, geocoded, and fed into the same matching pipeline as reports from the public.
Clustered map of lost, found, and adoptable pets with filters by species, status, source, colour, and date.
Upload a picture and get visually similar animals ranked by match score and distance.
Every listing shows its strongest possible matches, re-evaluated as new reports arrive.
Email alerts for new reports near a chosen address, and for new strong matches on your own pet.
A guided search checklist, printable posters, sharing, and saved pets for owners in the first critical hours.
API keys and rate limits for partner organizations, plus scheduled syncs and live operational stats.
Pawlert is the same work I do for clients, end to end: framing the real problem, choosing an architecture that will hold up, building it, and running it reliably in production. If your organization needs that combination of strategy and hands-on execution, let's talk.
Whether you need a Fractional CTO, Senior Technical Architect, technology advisor, or experienced technical leader for a critical initiative, let's talk.
Tell me what you're trying to accomplish, what's getting in the way, and where you need senior technical expertise.