Selected work/9 systems/Edge ↔ Cloud

Nine systems, and what each one taught me about failure.

Remote agents, cloud services, observability, update and recovery infrastructure, and software that runs inside cinemas. Every one of them exists because something upstream is unreliable, and pretending otherwise doesn't scale.

Environment
Edge ↔ Cloud
Primary constraint
Failure is the default
Method
Observe · Verify · Recover
01

Distributed systems

Edge agent platform

Go · SQLite · Distributed queues · Linux / Windows

Qube Cinema · Staff scope

Modernizing a long-lived remote agent into Go: durable local queues, cloud command handling, heartbeats, scheduling, and a behavior-preserving migration off a legacy runtime that nobody downstream is supposed to notice.

Engineering focus — Delivery semantics, restart safety, and clear ownership across concurrent workers.

02

Platform engineering

Safe update and recovery system

Go · A/B slots · systemd · Windows SCM

Qube Cinema · Staff scope

A self-update architecture for software deployed to machines I will never physically touch: verified artifacts, A/B installations, atomic activation, candidate probation, crash-loop detection, and automatic rollback.

Engineering focus — The known-good version stays recoverable through every step of the transaction.

03

Cloud & reliability

End-to-end observability

OpenTelemetry · OpenObserve · Kubernetes · OTLP

Qube Cinema · Staff scope

Instrumenting agent workflows across callbacks, schedules, persistent queues, and cloud commands with traces, structured logs, and metrics — then proving it on real virtual machines rather than in a diagram.

Engineering focus — Trace context survives the three places it usually dies: the queue, the restart, and the network hop.

04

Media systems · Rust

Next-generation cinema playback platform

Rust · wgpu · GPU rendering · DCI / SMPTE

Qube Cinema · Earlier role

Led the design and build of a next-generation playback system in Rust, with a wgpu-powered rendering pipeline for hardware-accelerated video. Owned the low-level and high-level design: service boundaries, security model, and a Rust-centric architecture.

Engineering focus — Frame stability and low latency, in a room where a crash is three hundred people in the dark.

05

Media systems · Rust

Encrypted GStreamer pipeline components

Rust · GStreamer · Encrypted media · Pipeline debugging

Qube Cinema · Earlier role

Built a set of custom GStreamer plugins in Rust for secure video and audio pipelines — encrypted sources, decryptors, and media processing elements — replacing fragile pieces of the runtime with ones that hold up under production load.

Engineering focus — Runtime stability across GPU rendering, encrypted streams, and real-time playback.

06

Reliability & compliance

Secure playback logging and operator tooling

Rust · Binary formats · XML · Web tooling

Qube Cinema · Earlier role

A tamper-resistant logging system in Rust using a binary log format with XML export, for forensic auditing and compliance. Then the tools that make it usable: a browser-based viewer that turns encrypted logs into something a human can read, and a web playback controller for managing and monitoring secure playback remotely.

Engineering focus — Evidence that holds up when somebody asks “are you certain?” in a distinctly legal tone.

07

Systems architecture

Safe Rust bindings for cinema libraries

C / C++ · Rust FFI · libasdcp · xmlsec

Qube Cinema · Earlier role

Developed and maintained safe Rust bindings for the C and C++ libraries the industry runs on — libasdcp and xmlsec — so cryptographic and DCP workflows get memory safety without discarding decades of proven domain behavior.

Engineering focus — Reduce the unsafe surface area; keep everything the old code already knows.

08

Migration · Rust

asdcplib, ported to Rust

Rust · C++ · MXF / KLV · SMPTE

Qube Cinema · Internal

Rewrote asdcplib — CineCert’s C++ AS-DCP implementation, the library the industry uses to pack JPEG2000 video, PCM audio, and timed text into MXF containers — as a Rust library, driven by the porting skill I wrote for coding agents. The core MXF reading logic compiled and ran correctly on the first pass. Test coverage and performance tuning are ongoing; it is an internal tool and not open source yet.

Engineering focus — Wrapping a C++ library buys safety at the boundary. Porting it moves the safety inside — which means redesigning ownership, not translating syntax.

How it was done →
09

Earlier work · C++

Backend systems and security review

C++ · Design patterns · Code review · Security

Wipro · 2021 – 2023

Back-end logic in modern C++ — templates and object-oriented design patterns for modularity and maintainability — plus sustained code review and analysis that found and closed more than 500 vulnerabilities across critical systems.

Engineering focus — The habit of reading code adversarially, which never stopped being useful.

The common thread

Make difficult systems boring.

Boring is the highest compliment a distributed system can receive. Languages change, platforms change, the fashionable architecture changes twice a year. The work doesn't: define contracts, expose failure, preserve evidence, and make recovery something the system does rather than something a person remembers to do.

More about me →