Engineering manager for HPC & AI cloud software at HPE, based in Minnesota. Day job is teams, platforms, and shipping infrastructure at scale. Outside of that I still write production-shaped systems — especially options trading software and evidence-backed local tooling.
LinkedIn · GitHub @endthestart
Senex Trader is a Django + TastyTrade platform: strategy factory (spreads through a custom 6-leg “Senex Trident”), DXLink streaming into the app, Celery automation, reconciliation, and risk controls. I’ve been writing about the architecture in a LinkedIn series (strategy engine, streaming, tooling tradeoffs).
The computational pieces that don’t need a broker live as stdlib-only MIT libraries extracted from that system:
| Library | What it actually does |
|---|---|
| optionstruct | Multi-leg structures → exact expiration risk (max loss/profit, breakevens); BS greeks; no I/O |
| strikemetrics | Chain positioning: GEX, max pain, strike walls, PCR, delta-OI |
| bookpolicy | Score open / close / hold against a book and target (PnL-blind by construction) |
| pivotstructure | Swing pivots → HH/HL/LH/LL structure states from a price series |
- local-llm-coding-evidence — measured local coding-agent routes on a 2019 Mac Pro (split dual-GPU) and 32 GB M2 Max: model × harness × placement × task, including failures and Wilson intervals. Not a leaderboard — a deployment guide.
- new-project-template —
repository-first scaffold: vision, ADRs, specs, and a harness-neutral
AGENTS.mdcontract. - backup-archeology — plan-first cleanup for backup/NAS trees (safe vs review tiers, quarantine default).
- vitamink — Rust/KDE helper that flips a dummy plug + Sunshine when the desk display sleeps (daemon exists; polish still in progress).
- plane-runway — Python CLI for self-hosted Plane (cycles, modules, local config).
I care about boundaries: stdlib cores vs broker adapters, immutable verifiers outside agent workspaces, and infra that’s boring enough to operate.