Two products born from real client engagements. Below them, a sample of software company work with anonymized outcomes.
A rubber-roll fabricator was losing margin on hand-planned cuts and burning operator hours on paperwork. We embedded for six months. We left them with a system.
Rubberfit is a Rust-powered nesting engine, an AI-augmented planner, audit-ready barcode lineage from receiving through shipping, and an operator-grade UI that runs the floor — designed and built by Arc & Anchor end to end.
Deep-AI code review. An ensemble of model specialists examines every pull request, debates the changes, and issues a single sealed verdict ready for the merge button. Built by Arc & Anchor for engineering teams that ship fast and review hard.
Visit Sigilix.aiA live two-pipeline trading system — prediction markets (Kalshi, Polymarket) and US equities (Alpaca) — running on Vercel cron with a Python research worker on GitHub Actions. A five-model ensemble (Claude, GPT-4o, Gemini, momentum, technical) feeds Bayesian log-odds combining and Kelly-sized risk management.
We built it to stress-test our own assumptions on cash-flow models, deal scenarios, and sensitivity analyses. It is not a public product— we don't sell signals, manage outside money, or offer trading advice. Available to clients only as part of a Finance engagement.
Internal use onlyAnonymized. Outcomes-first. Every engagement lasted at least a quarter; most ran six months or more.
We take on a small number of build engagements every year. If your problem looks like the ones above, we should talk.
Start a conversation