Selected work

Things I've built, broken open,
and put back together.

A mix of products I designed end-to-end and platform work inside large systems. Each one is a problem → approach → result, because that's what actually tells you whether I can solve yours.

BidHound

BidHound

AI-assisted freelance bidding · Android
product
Kotlin · Jetpack Compose · Spring Boot · Claude API
problem
Freelancers burn hours writing proposals; most are generic and lose.
build
An Android app that reads a job post and drafts a tailored bid via the Claude API, on a Spring Boot backend with a reusable auth module shared across projects.
result
Proposal drafting cut from minutes to seconds, with a shared platform-stack repo powering future apps.
NearMe

NearMe

Location intelligence · full-stack
product
Spring Boot · PostGIS · Android/Compose · Redis Streams
problem
Nearby-place search plus crowd-sourced price reporting needs fast geo queries and a way to spot bad data.
build
PostGIS-backed nearby search, a Room-cached Android client, a Redis-Streams error pipeline, and a domain-agnostic statistical outlier detector for price reports.
result
A modular platform where new place categories and detectors drop in without reworking the core.
100M+ / day

100M+ events/day pipeline

Real-time platform · media scale
platform
Kafka · Confluent Cloud · Kafka Streams · AWS ECS
problem
A real-time event platform needed to ingest and process well over 100M events a day, reliably.
build
Designed and ran the streaming pipeline end to end — producers, stream processing, sink connectors, and the Datadog observability around it.
result
Sustained nine-figure daily throughput with consumer-lag alerting that caught issues before they reached users.
v5 → v7

Zero-downtime Confluent migration

v5.3.3 → v7.2.0 · 10+ environments
rescue
Confluent · Terraform · AWS · cluster linking
problem
A major version jump everyone was nervous to attempt, across every environment, with zero tolerance for an outage.
build
Staged the upgrade with a tested rollback runbook, cluster linking, and Terraform-managed changes applied environment by environment.
result
Completed across 10+ environments with zero downtime — plus a fleet-wide RCA that fixed 50+ services via one IaC change.
50+ services

Fleet-wide failure, one fix

ECS state-lock RCA · infrastructure-as-code
rescue
AWS ECS · Kafka Streams · Terraform · Datadog
problem
A state-directory lock collision was taking down services fleet-wide, triggered by deployment overlap and disk exhaustion.
build
Traced the root cause across the fleet, then corrected task placement and bind-mount paths through one Terraform change applied everywhere.
result
50+ services stabilized from a single infrastructure-as-code fix, with the failure mode designed out for good.
Observability

Observability from scratch

Metrics · tracing · alerting
platform
Datadog · Grafana · OpenTelemetry · OpsGenie
problem
Teams were flying blind — no end-to-end visibility into a high-throughput streaming platform.
build
Built dashboards, distributed tracing, and consumer-lag monitoring from zero, with alert routing to on-call.
result
Incidents surface before users feel them, with a single pane of glass across brokers, topics, and services.
Claims Clerk

Claims Clerk

On-device scanning · firewalled data backend
product
Android/Compose · Spring Boot · on-device OCR · data boundaries
problem
Claims processing needs everyday tooling, but the documents are exactly the data you can't afford to leak.
build
A clerk app that scans and parses claim documents entirely on-device, with a system-of-record backend kept separate from a read-only automation path.
result
Sensitive images never leave the device to be read, and every access path has exactly the privilege it should. Read the case study →

More in the pipeline

New products and platform engagements are added here as they ship. Have one that belongs on this list? It could be yours next.

Start a project →

Want the same kind of result?

Tell me where your system is today and where it needs to be.

Request a project →