Software engineer · Dallas, TX
Cole McIntoshAI systems, without limits.
I build AI agents and extraction systems.
Selected impact
Evidence behind the work.
Each number is tied to a public project or contribution history.
Currently
Work
- Open Lease AuditAuditStreaming cross-document auditNext.js · TypeScript · Vercel AI SDK
- pdfmdRust~2,200 pages/secRust · PDF · CLI
- OpenExtractFramework11 model providersPython · Pydantic AI · Multimodal
- NudgeAgentDurable case workflowsPython · Deep Agents · Human in the loop
- LangChain SalesforceIntegrationPublished on PyPIPython · LangChain · Salesforce
- CoT RerankingRetrievalReasoning optimizationPython · Cohere · Llama
Experience
OpenExtract (PyPI): typed extraction across 11 LLM providers.
LLM document extraction for delinquency operations.
Internal automation for the ghost kitchen network.
Reporting infrastructure for property and portfolio metrics.
Stack
Open source
For teams
I help development teams redesign their software lifecycle around AI coding tools — so agents add real leverage, not noise.
-50%
cycle time
I've helped cut feature lead time by handing mechanical work to agents and keeping humans on the decisions that ship value.
~10h
eng / wk
I've helped reclaim roughly a day a week per engineer from boilerplate, migrations, and review toil — capacity that shows up on the roadmap.
+40%
to prod
I've helped push more agent-authored code through CI and into production, with evals catching the slop before customers feel it.
2×
PR volume
I've helped teams increase pull-request throughput with agents in the review loop — without lowering the bar.
-35%
rev wait
I've helped shrink review queues with agent-assisted first passes that flag risk before a human opens the diff.
3×
agent use
I've helped orgs go from a few power users to Cursor, Codex, Claude Code, Amp, Factory Droids, and OpenCode as default.
Cursor Agent, Claude Code, OpenAI Codex, Amp, Factory Droids, and OpenCode — which harness fits the team, and which to skip.
Reviews, prompts, evals, and guardrails woven into the loop you already run.
MCP, CI, and wiring agents into the repos and services you ship with.
Contact
I work with teams building agentic products, document extraction, and AI-enabled engineering workflows.
cole@staymellow.ai