I design systems that ship.
Document. Repeat.
Three production-grade projects built in under a month using AI as an engineering partner. Each one tested a different aspect of working with AI — specification, architecture, evaluation, and the domain expertise only a human can bring.
Each project raised the stakes — more complexity, tighter integration, harder problems. The sequence was deliberate: learn the workflow on a Python pipeline, apply it to a Swift monorepo, then push it to a full product with AI generation, subscriptions, and a backend.
Codeword Pipeline
6-stage real-time ML pipeline. Monitors live audio, transcribes, classifies, extracts structured data. Running in production with 100% accuracy.
BLEKit
Two-app monorepo with a shared BLE scanning engine. SentinelScan for security audits, Overwatch for counter-surveillance. 49 files, 11,500 lines.
CipherRank
Gamified Security+, Network+, and SecAI+ prep with AI mission generation, two-stage validation, cost-modelled token economics, and StoreKit 2 subscriptions. 648 missions, 2,242 decision points. In App Store review.
Codeword Pipeline
6-stage real-time ML pipeline. Monitors live audio, transcribes, classifies, extracts. 100% production accuracy.
BLEKit
Two-app monorepo with shared BLE scanning engine. 49 files, 11,500 lines.
CipherRank
Gamified Security+, Network+, and SecAI+ prep with AI generation pipeline, two-stage validation, and cost-modelled economics. 648 missions. In App Store review.
Each project page is a deep dive into how the system was designed, what decisions mattered, and what I'd do differently. Not a feature list — a thinking record.
Codeword Pipeline
A 6-stage real-time ML pipeline that monitors live audio, classifies speech, and extracts structured data. Running in production.
BLEKit
Two iOS security apps sharing a common BLE scanning engine. Site audits for professionals, counter-surveillance for journalists.
CipherRank
Gamified CompTIA exam prep — Security+, Network+, and SecAI+ — with AI mission generation, two-stage validation, RPG progression, and cost-modelled token economics.
Longer-form thinking on AI systems design, security automation, and the decisions behind the projects. Expanding on ideas that start as LinkedIn posts.
Everything on this site was designed and built with AI as an engineering partner — not a code generator, not an autocomplete. The AI handled syntax, pattern application, and the mechanical translation of intent into code. I handled architecture, specification, product decisions, domain expertise, quality evaluation, and every decision that required understanding what the system is for, not just how it works.
That split between human judgment and AI execution is not a workaround for lacking traditional credentials. It is the emerging model for how software gets built — and the skill is knowing what to build, specifying it precisely enough that an AI can execute it, and evaluating whether the output meets the bar.