Skip to content
View klmn800's full-sized avatar

Block or report klmn800

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
klmn800/README.md

Ben Keilman

Boston-based builder of internal AI tools - and the person who makes sure they actually get used.

I find what's broken in how a team works, build an AI-assisted tool to fix it, and drive the adoption that makes it stick. Operations-analyst judgment, three years building on the Anthropic API (RAG, agents, automation), and a background in training and enablement - so the tool ships and people use it.

How I work

I work with Claude Code as a co-builder. That compresses "I have an idea" to "the tool exists and runs against real data" from weeks to hours - whether it's a one-afternoon utility or a system that grows over months. Everything pinned below was built this way: solo, end-to-end, against problems I or my colleagues actually had.

What's here

  • policy-navigator - My biggest RAG build: a multi-source RAG system over U.S. public-assistance (SNAP) policy. It indexes four authoritative sources - federal program regulations, the federal quality-control standard, a state's regulations, and that agency's ~1,200-page internal staff guide (10,000+ searchable chunks in a shared vector store) - so staff can ask a policy question and get a synthesized, cited answer. It also cross-checks the sources against each other to surface gaps and conflicts, drafts new policy pages from research, and analyzes quality-control error findings. FastAPI + HTMX web UI; Claude API for structured synthesis, Claude Code CLI for the agentic, resumable work. Proof the AI-assisted approach scales well past small tools.
  • options_scanner - My longest-running system, now public: a personal options-flow research platform that has run unattended every trading day since 2025. Real-time flow monitoring over a ~820-symbol universe, open-interest time series, earnings intelligence, and a SQLite datalake whose main scan table holds 23M+ rows. It files its own bug reports into Claude Code sessions (autofix), and five autonomous Claude Code agents audit its data quality and research trades overnight - the agent framework is published separately as agent_lab. Shared as a read-only showcase.
  • doc-review-pipeline - Batch document reviewer that pulls .docx attachments out of emails (.msg/.eml), runs them through Claude, and writes structured analysis to a tracker. Built to replace a manual review queue.
  • query-db-direct - SQLite exploration tool with schema discovery and dynamic query suggestion. Designed so AI agents can use it as readily as humans can.
  • media-wall - Local masonry-grid media viewer with autoplay, tagging, and filtering. A self-hosted alternative to cloud galleries.
  • archive-cracker - Utility for recovering personal media from locked .rar archives.

More in progress.

Background

Business Operations Analyst by day. The options_scanner system above is my own - a 100GB+ market-analytics datalake with a multi-agent system, built and operated solo as a live production system for over a year. I also run a freelance practice partnering with domain experts who have deep knowledge but limited engineering capacity (most recently a building-code RAG for an architecture firm). I'm used to translating between business problems and technical systems, scoping the work, shipping the fix, and making sure it gets adopted.

What I'm looking for

Roles where I build the internal tooling an organization needs and help people adopt it - AI implementation and enablement, internal tools and automation, or applied AI engineering. Boston - in-office or hybrid preferred, open to remote.

Reach me

LinkedIn

Pinned Loading

  1. policy-navigator policy-navigator Public

    Multi-source RAG over SNAP public-assistance policy: semantic search with citations, cross-source gap detection, and AI-assisted policy authoring.

    Python 1

  2. options_scanner options_scanner Public

    Personal options-flow research system: flow monitoring, open-interest analysis, earnings intelligence over a ~820-symbol universe. Published as a read-only showcase.

    Python 1

  3. os-agent-earnings-researcher os-agent-earnings-researcher Public

    Options Scanner — Earnings Researcher agent (pre-earnings dossiers)

    HTML

  4. claude-code-runner claude-code-runner Public

    Python wrapper that drives the Claude Code CLI as a subprocess. Spawn agentic sessions, capture structured results, hand sessions to humans. Single file, standard library only.

    Python 2

  5. doc-review-pipeline doc-review-pipeline Public

    A configurable batch document reviewer powered by Claude. Drop Word documents (or Outlook emails containing Word attachments) into a folder, and the pipeline extracts the text, sends each document …

    Python 1

  6. claude-autofix claude-autofix Public

    Python