cooper anderson

AI Agent Infrastructure & Full-Stack Engineer

San Francisco, CA email github linkedin resume.pdf ↓

Cooper Anderson

Builds AI-agent infrastructure -- multi-agent orchestration, MCP servers, and persistent Rust and Python daemons -- and ships production full-stack SaaS that holds up under real load.

Frontend engineer at Veoci; co-lead of a ~143K-line options-flow analytics platform through my LLC. Rust where it has to be fast, TypeScript where it has to ship, Python where the data lives.

  • 143K loc platform co-led
  • 5 mcp servers built
  • 39s -> <1s api route rewrite
  • 6 languages in production

contribution ledger

— contributions in the last year

source: github.com/mcande21 — live

experience

Software Engineer @ Veoci — Greywall Software

Jul 2025 -- Present

Remote -- Vue.js / TypeScript web client of an enterprise emergency-management SaaS.

  • Core contributor since inception to the team's internal AI-agent orchestration framework -- shipped a Bugsnag-triage agent skill (~1.8K LOC), LSP auto-integration, and a war-room multi-agent collab system (Rust CLIs); post-adoption, monthly merged PRs rose ~35% (~360 to 490+) with no added headcount.
  • Sole author of the framework's second generation -- tailored to developer workflow: session continuation, per-developer persistent memory, a hybrid wiki + Neo4j/Graphiti graph-memory tier over MCP, and a self-improvement loop that captures corrections and evolves its own instructions.
  • Shipped the Blueprints reusable-template system (~16.7K LOC) for the "Pages" no-code page-builder -- recursive block-tree editor, 45+ block types, Layers/outline panel, and real-time multi-user presence.
  • Diagnosed a Vue reactivity loop in the Leaflet layer that crashed 599 production events; shipped the fix with regression coverage.
~35% Team PR throughput

stack Vue 3, Vuex, TypeScript, Leaflet, Firebase, Froala, Storybook/Jest, i18n

Independent Software Engineer @ CASystems LLC

Nov 2025 -- Present

New York -- Single-member NY LLC for product and consulting engineering.

  • OptionsFlow (co-lead) -- co-built a production options-flow data & analytics SaaS (Next.js 16 / React 19 / Neon Postgres / Redis / Stripe; ~143K LOC). Owned the trade-ingestion pipeline and trades hot-path: re-architected ingestion for ~10x throughput (~900 to 9,000+ trades/day) and cut an ~11s query regression via a Postgres partial index + read replica.
  • The Daily Traders / TDT (contract, Rumford Capital) -- rebuilt the member-portal app in Next.js 14: cut a core signals API route from ~39s to <1s via a bulk match-context builder, built a GEX enrichment engine, and ported ThinkScript ADR take-profit/stop-loss logic to TypeScript (282 tests).
  • AJ's Airport Runs (client) -- shipped a production ride-share platform: a Vue 3 + Capacitor iOS/Android app (~73K LOC, 25+ Firebase Cloud Functions, pricing engine, Stripe, signed App Store v1.1.0 build), then rebuilt it in React Native / Expo on Supabase/PostGIS.
10x Ingestion throughput

stack Next.js 16, React 19, Neon Postgres, Redis, Stripe, React Native/Expo, Supabase/PostGIS

selected work

simple-subs

source ↗

A subagent dispatch library that coordinates single, parallel, and async agent execution with peer messaging. Published on npm. Features concurrency control, filesystem inbox protocol, model routing, and a compact TUI. The kind of library you write when you realize the existing tools are too heavy for the job.

TypeScriptnpm

openclaw-bridge

source ↗

A CLI bridge with three modes: raw CLI, ratatui TUI, and MCP Channel Server. WebSocket transport with Ed25519 authentication over Tailscale networking. Distributed via Homebrew tap. Systems programming where it matters.

RustCLI

ulster-elections

source ↗

"Brian's Boards 2025" -- competitiveness analysis, PDF upload, multi-county support. Full-stack with FastAPI backend and TypeScript frontend. Civic tech that made it to production.

PythonFastAPI

audio-duplicates

source ↗

An npm library using Chromaprint fingerprinting with a native C++ implementation. Multi-threaded scanning, TypeScript definitions, CLI tool. When JavaScript isn't fast enough, you drop to the metal.

C++npm

signal-dispatch

source ↗

A geopolitical intelligence newsletter with its own blog and marketing site. Research pipeline processes global signals into structured analysis. Because the news doesn't cover what actually matters.

PythonResearch

client & private work

Persona-based multi-agent review system

Details under NDA.

Rust + Python -- active development -- 2026

Three-repo architecture: a Rust orchestration core dispatching persona-based agents, a knowledge-graph memory server built on Graphiti MCP, and a structured reference corpus. Agents assume specialized roles, pull context from the graph over MCP, and evaluate inputs against a codified rule framework -- the reasoning is grounded in retrieved structure, not model recall.

  • Rust orchestration core with a Python tooling layer
  • Persona-based agent dispatch with domain specialization
  • Temporal knowledge-graph memory tier served over the MCP protocol
  • Corpus ingestion pipeline with structured retrieval
RustPythonAI Agents

Provenance-gated drafting engine

Details under NDA.

Python -- active development -- 2026

Document pipeline built on one doctrine: the LLM reads, deterministic code writes. Models do the fuzzy work -- extracting facts from hostile scanned sources -- while checkable code owns assembly, templating, and consistency. A contamination gate fails closed on any generated fact it cannot trace to a source page, and no model asserts a fact from memory: everything checkable lives in an MCP tool server, not the model. A human reviewer signs off on every output.

  • Every generated fact carries a page-level citation to its source
  • Contamination block-gate: unsourced assertions fail closed
  • Deterministic core proven by 213 tests; validated end-to-end on real source material, locally, air-gapped
  • MCP tool server owns document assembly, templating, and consistency checks
PythonAI AgentsMCP

Low-latency member platform rebuild

Details under NDA.

TypeScript -- production -- 2025-present

Contract rebuild of a production member platform in Next.js, serving a live paying subscriber base. The hot path: a bulk match-context builder took the core API route from ~39s to under a second. Companion services include a real-time data-ingest pipeline, an enrichment engine layered on the ingested stream, and bots that format and deliver alerts to the community.

  • Core API route cut from ~39s to <1s via a bulk context builder
  • Real-time data-ingest service feeding an enrichment engine
  • Proprietary strategy logic ported from a trading DSL to TypeScript (282 tests)
  • Alert-delivery messaging bots in production
TypeScriptNext.jsSaaS

Automated research pipeline

Private repository.

Python + TypeScript -- active development -- 2026

Research pipeline that normalizes source data into a structured store, tracks probabilities as time series, and feeds two clients off the same data layer: a terminal TUI for live monitoring and a web dashboard for visualization.

  • Automated collection runs with structured normalization
  • Time-series probability tracking across sources
  • Terminal TUI and web dashboard reading one shared data layer
PythonTypeScriptResearch

Sandboxed AI evaluation engine

Details under NDA.

TypeScript -- development -- 2026

Evaluation platform with real-time model integration: submitted inputs are scored against rubric criteria through an analysis pipeline that generates structured feedback. Untrusted code never touches the host -- execution is isolated in microVMs, and the pipeline treats everything it grades as hostile.

  • AI-powered evaluation pipeline with structured output
  • Untrusted code execution isolated in microVMs
  • Criteria-based scoring and feedback generation
TypeScriptAI

reference desk

Still have questions? Ask the record. A language model answers from this site's data — it cites its sources, speaks in the third person, and refuses what it cannot source.

generated from the record by a language model — verify against the page, or email the correspondent. also answers from the terminal: curl -X POST /api/ask