Senior Full-Stack Engineer — AI & agentic systems, GIS & mapping, real estate tech.
Most of my work lives in private repositories. This is a look at what I've shipped.
"Because the most important part of any deal is making it home safe."
tetherre.com · App Store · Google Play · Flutter · Web
Real estate agents meet strangers alone, in empty houses, on unfamiliar streets. Tether RE is a personal safety platform built for that reality — continuous monitoring, one-touch emergency dispatch, and client screening that runs from first contact through close.
As Head of Engineering, I took over the entire engineering function as the company consolidated to a single engineer, and owned the product from a B2C app through a full pivot to B2B SaaS — reworking the codebase, pricing, and go-to-market.
| Growth | 0 → $40K MRR · 5,000 users · multiple enterprise contracts |
| Recognition | Won the 2024 T3 Technology Summit Pitch Battle · selected for the NAR REACH Accelerator |
| Platforms | Flutter with native iOS & Android modules · web dashboard |
24/7 live monitoring. Professional dispatchers watch sessions in real time and can call EMS even when the agent is unable to speak or reach their phone.
SOS and silent dispatch. A tap discreetly notifies emergency contacts. A press-and-hold skips monitoring entirely and dispatches help immediately — built for situations that are already escalating.
Struggle and impact detection. On-device sensor analysis flags falls, assaults, and car crashes, then opens a safety protocol automatically.
Proximity safety timers. If an agent stays at a property longer than expected, the app checks in on its own.
Real-time GPS tracking. Continuous location, streamed to monitoring for the duration of a showing.
Engineering notes — I built the app's real-time location layer: live GPS tracking, proximity-based safety alarms, and group navigation that keeps an agent and their clients synced on the same route in real time. The mapping and routing underneath runs on a custom Flutter + MapLibre navigation package I wrote after migrating off Mapbox — which cut mapping and routing costs ~90%.
Screening happens before the agent ever gets in a car.
- Reverse phone number lookup
- Reverse address lookup
- Criminal background checks
- Sex offender registry checks
Every agent gets 12 free lookups a month.
Safety gets agents to install the app. These features get them to keep it open.
- Unlimited auto-logged mileage tracking
- AI-powered expense tracking
- Turn-by-turn navigation
- Buyer notes and showing logs
- Showing organizer and dashboard
- Custom branded experience per brokerage
Engineering notes — I designed and implemented the AI receipt parsing behind expense tracking on AWS Bedrock, so agents can snap a receipt and have the expense logged automatically.
Brokerage and association administrators manage rosters, branding, and safety reporting from the browser.
Engineering notes — Alongside the COO I architected the company's core business systems: StaxBill subscription & enterprise billing with org-hierarchy revenue reporting, HubSpot tooling for account management, and UserPilot for product analytics. I also built AI-powered analytics dashboards that unified billing, user analytics, and revenue into a single view for leadership.
"Moves worth making, drafted while you slept."
realrev.ai · Agentic AI · Python · Multi-tenant SaaS
RealRev is an agentic AI platform that works an agent's business while they sleep. Overnight, a fleet of autonomous agents scans the open web, listing and MLS data, and social platforms — then surfaces the specific moves worth making and drafts the outreach to make them. Everything waits for the agent's approval; nothing sends until they say so.
I designed and built RealRev end to end — the agent orchestration, the scheduled always-on jobs, the OSINT scanning layer, and the human-in-the-loop review that keeps a person in control of everything that goes out.
Each morning the agent wakes up to a prioritized queue of opportunities — a fresh price reduction paired with national press, a listing worth a marketing push — each scored, explained ("why this matters"), and packaged with drafted email and social content ready to send with one tap.
Autonomous agents run on their own schedules — morning briefings, lead radar scans, strategic analysis, and business OSINT — quietly working in the background between logins.
RealRev continuously audits an agent's digital footprint and reputation across the web — catching broken listing syndication, dual-MLS confusion, and name/brand inconsistencies before they cost real buyer exposure, and tracking mentions and profiles across LinkedIn, Facebook, and Instagram.
RealRev is a large Python system I designed and built end to end.
- Three-tier agent architecture on the Anthropic SDK — a manager agent routes to department routers, which dispatch sub-agents running tool-use loops over ~117 tool definitions, with a skill-and-capability registry resolving to per-agent tool grants. Tool failures get their own channel, so an agent can't narrate a success it didn't actually achieve.
- Multi-tenant provisioning that stands up a customer end to end — host assignment, DNS, Twilio subaccount and number, container deploy over SSH, license issuance, and first-boot config — as resumable step machines with idempotent retries across five external APIs.
- Multi-provider model routing across OpenRouter, Fireworks AI, and local Ollama behind an operator-controlled role map, cutting inference spend ~60%, with prompt caching and per-call cost metering throughout.
- Python 3.12 · FastAPI · async throughout — raw-SQL SQLite per tenant (WAL, FTS5) and PostgreSQL via asyncpg. Deny-by-default auth, HMAC webhook verification, SSRF guards, and injection-safe column whitelisting.
- Docker · Traefik with host-header routing, Cloudflare Tunnels and Zero Trust, DigitalOcean, and GitHub Actions CI/CD with digest-pinned images. ~230 tests with the network blocked at the transport layer.
Stack: Python 3.12 · FastAPI · Astro · Anthropic SDK · asyncpg / SQLite (WAL, FTS5) · Docker · Traefik · Cloudflare Zero Trust · Twilio · DigitalOcean · GitHub Actions
At EagleView I built GIS-powered web applications across the imagery platform — both Cloud Explorer (viewing, measurement, and analysis of high-resolution aerial imagery) and EagleView 3D (digital-twin mesh models for 3D measurement, line of sight, and shadow analysis) — on a customized fork of Mapbox.
- Bridged two geospatial ecosystems — converted ArcGIS tile data into Mapbox-compatible tiles so existing imagery could drive the new platform.
- Modernized a legacy GIS platform by migrating core services to a scalable cloud architecture, and integrated third-party 3D modeling to extend what enterprise customers could analyze.
- Built a life-safety feature that lets 911 dispatchers determine a caller's elevation inside a building during emergency calls.
- Defined API contracts across product and data science, and implemented Okta SSO for third-party integrations.
Stack: Mapbox · ArcGIS · PostGIS · cloud GIS services · Okta SSO
At Thirtythree I built a geospatial data-collection platform for the 5G network rollout that maps and identifies cell-tower sites, then turns that field data into the drawings used for permitting and construction. It pairs a Site Selector mobile app with a Control Tower web dashboard, and — per Thirtythree — lets crews complete deployments up to 10× faster.
- Full-stack across mobile, mapping, and data output — field data capture on mobile, the dynamic mapping layer (street / satellite / 360°), and automated report generation (Word / Excel / PDF), delivered in React, React Native, Node.js, and Go.
- Instant GPS coordinates, automatic address lookups, and distance calculations to cut manual data entry and get sites captured right on the first pass.
In testing, the platform ran full site assessments in under 10 minutes and processed 1,500 candidate SCIPs in under 10 days. · Stack: React · React Native · Node.js · Go
| Layer | Technology |
|---|---|
| Mobile | Flutter · Dart · native iOS (Swift) · native Android (Kotlin / Java) |
| Mapping | MapLibre · Mapbox · custom Flutter navigation & routing package |
| Web | React · Next.js · Astro |
| Backend | Python 3.12 · FastAPI · Node.js · NestJS · gRPC · Protocol Buffers |
| AI | Anthropic SDK · AWS Bedrock · multi-agent orchestration · multi-provider model routing · RAG |
| Data | PostgreSQL · PostGIS · asyncpg · SQLite (WAL, FTS5) · Firebase / Firestore |
| Business systems | StaxBill · HubSpot · UserPilot · Twilio |
| Infrastructure | AWS · DigitalOcean · Docker · Traefik · Cloudflare Zero Trust · GitHub Actions · GitLab |
My overall stack across RealRev, Tether RE, EagleView, and other work.
- Email — brandub@gmail.com
- LinkedIn — brandoncodes
- GitHub — @brandub