Skip to content

Projects

Engineering work in context

Open-source tools, product work, and anonymized production systems. Each case study focuses on the problem, the constraints, the decisions that mattered, and the lessons that carried forward.

Featured

Public projects I can discuss in full, from the problem they solve to the architecture and trade-offs behind them.

  • Reliability InfrastructureOpen sourceActive

    RelPrim

    Creator & Maintainer

    RelPrim is an open-source Python reliability layer for operations that cross process, network, or provider boundaries. It brings retries, timeouts, fallbacks, circuit breakers, validation, and execution reporting into one explicit operation model without hiding the domain decisions that make recovery safe.

    • Python
    • Reliability policies
    • Structured execution
  • AI Creative PlatformActive

    Ghostviber

    Founder & Engineer

    Ghostviber is an AI creative platform built around the full songwriting workflow, from the first line to generated audio and artwork. I designed and built the product architecture, multilingual rhyme engine, AI orchestration, credit ledger, asynchronous workflows, storage, and reliability mechanisms needed to turn variable external providers into one coherent user experience.

    • AI orchestration
    • Distributed workflows
    • Usage-based billing

Production work

Professional case studies

Anonymized accounts of production work where reliability, scale, and operational constraints shaped the solution. Client names and confidential details are omitted, while the engineering problems, trade-offs, and lessons remain.

  • Payments & IntegrationsPast Work

    Making Payment Workflows Reliable

    Payment webhooks can arrive late, more than once, or after an ambiguous timeout. This case study explores how durable ingestion, idempotent processing, and explicit recovery paths protect financial correctness at the boundary with external providers.

    • Idempotency
    • Webhooks
    • Async processing
  • Connected SystemsPast Work

    Keeping Connected Security Systems Predictable

    Commands sent to connected security devices do not always produce immediate or certain outcomes. This case study examines explicit state transitions, asynchronous execution, auditability, and recovery when software actions affect the physical world.

    • Event-driven workflows
    • State transitions
    • Auditability
  • Ad Tech & Data PlatformsPast Work

    Turning Web Content Signals into Campaign Controls

    Raw web content is unreliable, constantly changing, and difficult to use directly. This case study shows how crawling, structured processing, and ML-backed classification were connected to practical campaign controls.

    • Web crawling
    • Data pipelines
    • ML integration

Let's talk systems

Building a system that has to remain predictable under pressure?

I help teams shape the systems their products depend on, especially early on or when reliability becomes critical. I also write and speak about these problems. Tell me what you’re building and where it hurts.