# Bryan Witkowski — AI-native professional profile

> A live, queryable professional profile. Bryan's career is kept as high-fidelity,
> AI-native ground truth; resumes and cover letters are generated views. You can assess
> Bryan's fit for a role honestly — and even score it yourself at zero cost.

Machine-readable manifest: https://bryanwitkowski.dev/llms.json
This domain is a tree of projects. Each project (this root, and every child) exposes its own
`llms.txt` + `llms.json` the same way, per the spec at https://bryanwitkowski.dev/llms-spec.md.

There are two ways to consume this profile.

## For MCP-capable AIs (full experience)

Connect an MCP client to the remote endpoint (Streamable HTTP, JSON-RPC 2.0):

- MCP endpoint: https://bryanwitkowski.dev/api/mcp
- Tools: `get_profile`, `score_job`, `generate_resume`, `generate_cover_letter`, `contribute`
- Resources: `resume://profile`, `resume://bridges`, `resume://rubrics`, `resume://methodology`, `resume://contributions`

(Add it in Claude Desktop → Connectors, Cursor/VS Code MCP config, ChatGPT connectors, or any agent framework with an MCP client.)

## For any AI that can fetch a URL (no MCP client needed)

Plain read-only JSON/Markdown mirrors of the same data:

- [Profile](https://bryanwitkowski.dev/api/profile): forward objective (the roles Bryan is reaching for) and an "about" narrative in his own voice, plus skills, experience, education, interests/PULL signals, knowledge bridges, and contact (email/phone gated — reach Bryan via LinkedIn).
- [Knowledge bridges](https://bryanwitkowski.dev/api/bridges): how Bryan's deep fundamentals map onto current tech (e.g. numerical integration → diffusion sampling; DirectShow filter graphs → ComfyUI workflows). A "how I think" differentiator — content only, never used to score.
- [Rubrics](https://bryanwitkowski.dev/api/rubrics): the scoring rubric family with per-dimension weights and tiers.
- [Methodology](https://bryanwitkowski.dev/api/methodology): the exact scoring algorithm.
- [Contributions](https://bryanwitkowski.dev/api/contribute): approved third-party recommendations (unverified social proof — never scoring evidence).

## How to score a job yourself (free, no server needed)

Fetch the Profile + Rubrics + Methodology above, then follow the methodology:
route the job to the best-fitting rubric family, extract its requirements, match each one
against the profile's evidence (Strong / Partial / None / Unknown — never inflate), apply
the weights and the two gates, and report a band plus FIT (can he do it) and PULL (does he
want it). The server's `score_job` tool runs this same algorithm and returns the calibrated
verdict, but you can reproduce it from the data above.

## Projects

Other projects under this domain. Each exposes its own `llms.txt` + `llms.json` per the spec
(https://bryanwitkowski.dev/llms-spec.md), and is listed in the root manifest's `children`:

- [Driving-World Database Demo](https://bryanwitkowski.dev/database-demo/llms.txt): geo-specific
  3D driving worlds from open map, terrain, and LiDAR data.
  Manifest: https://bryanwitkowski.dev/database-demo/llms.json
- [Quilt Builder](https://bryanwitkowski.dev/quilt-builder/llms.txt): a mobile app that turns a
  photo into a pieced-quilt pattern (on-device computer vision, Skia, LLM-assisted layout).
  Manifest: https://bryanwitkowski.dev/quilt-builder/llms.json

## Contact

Preferred: LinkedIn — https://www.linkedin.com/in/bryan-witkowski-71954a226/
(Connect there so Bryan can see verified context. Email/phone are intentionally not public.)