job_ops-mcp 0.3.0

package/.env.example

@@ -0,0 +1,33 @@
+# mcp-jsa configuration
+
+# Network
+MCP_JSA_PORT=7891
+MCP_JSA_HOST=127.0.0.1
+
+# Project root — where cv.md, config/profile.yml, portals.yml live.
+# Defaults to the directory this server runs from.
+# To reuse the existing career-ops/ data, point it there:
+# MCP_JSA_PROJECT_ROOT=/Users/you/Documents/Career Ops/career-ops
+MCP_JSA_PROJECT_ROOT=
+
+# Data + outputs (resolved relative to the server install dir, not project root)
+MCP_JSA_DATA_DIR=./data
+MCP_JSA_OUTPUT_DIR=./output
+
+# LLM provider — only used on api/batch paths in later milestones.
+# chat-mode (default) runs zero LLM calls server-side; the chat client is the brain.
+MCP_JSA_LLM_PROVIDER=gemini     # gemini | deepseek | none
+MCP_JSA_LLM_MODEL=
+GEMINI_API_KEY=
+DEEPSEEK_API_KEY=
+
+# Scheduler — off by default. Enable specific jobs via scheduler_enable MCP tool.
+MCP_JSA_SCHEDULER_ENABLED=false
+
+# Visa scoring — set to `false` if you don't need visa/H1B signal (US citizens, non-US
+# users, etc.). When false, visa_fit is dropped from the rubric, weights renormalize to
+# resume 0.6 / taste 0.4, the visa_signal / import_h1b / import_linkedin tools are hidden,
+# and visa columns are stripped from tool responses. Default: true.
+MCP_JSA_VISA_SCORING=true
+
+# Playwright — usually no config needed once `npm run playwright:install` is done.

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Mohith Kattukuttiyil Das
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/README.md

@@ -0,0 +1,400 @@
+# job_ops-mcp
+
+A self-hosted **Model Context Protocol** server for the full job-search loop — portal
+scanning, JD evaluation, tailored resume + cover PDFs, outreach drafting, story bank,
+negotiation brief — all driven from your MCP-aware chat client (Claude Desktop, Cursor,
+any client that speaks streamable-HTTP MCP).
+
+The chat is the brain. This server executes the mechanical work and hands every artifact
+back as an `http://localhost:7891/...` link.
+
+> **Status:** early. Works. APIs may still move pre-1.0.
+
+---
+
+## Quickstart
+
+```bash
+# 1. Scaffold your working directory (cv.md, profile.yml, portals.yml + SQLite DB)
+npx job_ops-mcp init
+
+# 2. Open cv.md, config/profile.yml, portals.yml and replace every <TODO> placeholder.
+
+# 3. Confirm everything is wired
+npx job_ops-mcp doctor
+
+# 4. Boot the server (Chromium auto-installs on first run)
+npx job_ops-mcp start
+#  ▷ job_ops-mcp listening on http://127.0.0.1:7891
+
+# 5. Get a copy-paste config block for your MCP client
+npx job_ops-mcp connect
+
+# 6. Paste a job URL or pasted JD into your chat — the chat calls evaluate_job, draws on
+#    your rubric + career_packet + tailoring rules, and walks the rest of the workflow.
+```
+
+That's the loop. Everything else (warm-intro finder, story bank, negotiation brief,
+batch rater, scheduler, …) is wired in but optional.
+
+---
+
+## What it does
+
+Two systems merged into one MCP server:
+
+- **Evaluation + materials side** — port of [santifer/career-ops](https://github.com/santifer/career-ops):
+  6-block A–F (+G legitimacy) report, archetype detection, ATS-friendly HTML→PDF resume
+  + cover generation, story bank, negotiation playbook.
+- **Pipeline side** — port of a personal Postgres + n8n pipeline ("JSA"): Greenhouse /
+  Ashby / Lever / Workday pollers + closed-board Playwright scrapers, content-hash
+  dedupe, batch LLM rater with strict-JSON parsing, warm-intro / founder DM drafter,
+  visa signal from DOL OFLC H1B data.
+
+Everything lives in a single Node process with a single SQLite file. No external
+Postgres, no n8n, no cloud anything. Bring your own LLM key (Gemini free tier by
+default, DeepSeek optional) if you want the API/batch paths; chat-mode tools work
+without one.
+
+> **Not affiliated with or endorsed by santifer's career-ops.** This is an independent
+> project that ports + adapts the publicly-released MIT-licensed templates and rubric
+> shape into the MCP transport surface. See the [Attribution](#attribution) section.
+
+---
+
+## Tools (36 — one MCP `tools/list` call away)
+
+| Group | Tools |
+|---|---|
+| **Evaluation** | `evaluate_job`, `batch_evaluate`, `get_top_jobs`, `evaluate_training`, `evaluate_project` |
+| **Materials** | `generate_materials`, `render_pdf`, `get_report` |
+| **Tracker** | `get_tracker`, `update_status`, `mark_ready_to_apply` |
+| **Sourcing** | `scan_portals` (Greenhouse + Ashby + Lever + Workday + Amazon + Google + generic Playwright) |
+| **Outreach** | `find_warm_intros`, `find_founders`, `draft_outreach`, `draft_followup`, `draft_reply`, `get_outreach_queue`, `update_outreach`, `get_followups_due` |
+| **Interview / offer** | `extract_stories`, `get_story_bank`, `negotiation_brief` |
+| **Research** | `deep_research`, `enrich_company`, `daily_digest` |
+| **Profile + ops** | `get_career_packet`, `update_career_packet`, `cost_estimate` |
+| **Apply (preview only — never submits)** | `apply_prefill` |
+| **Visa (optional, can be hidden)** | `visa_signal`, `import_h1b`, `import_linkedin` |
+| **Scheduler (opt-in cron, off by default)** | `scheduler_status`, `scheduler_enable`, `scheduler_disable` |
+
+Six MCP **resources** carry the editable behaviour — rubric, report_format,
+tailoring_rules, outreach_tone, negotiation_playbook, career_packet — all loaded from
+`modes/*.md` and live-reloaded on edit. Tune scoring or tone without touching code.
+
+---
+
+## Designed to be made yours
+
+The defaults assume nothing about your location, citizenship, role, or industry. Every
+behaviour-shaping piece is a markdown file you can rewrite:
+
+| You can change… | By editing… |
+|---|---|
+| Scoring dimensions + weights | `modes/rubric.md` |
+| 6-block report shape | `modes/report_format.md` |
+| Resume/cover tailoring rules | `modes/tailoring_rules.md` |
+| Outreach tone + char caps | `modes/outreach_tone.md` |
+| Negotiation scripts + framework | `modes/negotiation_playbook.md` |
+| Your bullet/project bank | `modes/career_packet.md` (or via `update_career_packet`) |
+| Tracked companies + filters | `portals.yml` |
+| Identity + target roles | `config/profile.yml` |
+
+### Non-US users / non-sponsorship cases
+
+Visa scoring is **fully optional**. Set:
+
+```bash
+export MCP_JSA_VISA_SCORING=false
+```
+
+When off:
+
+- `score_total = round(0.6 · resume_fit + 0.4 · taste_fit)` (server-side authoritative)
+- The `visa_signal`, `import_h1b`, `import_linkedin` tools are hidden from `tools/list`
+- `score_visa_fit` is stripped from `get_top_jobs` items and the eval-report HTML badge
+- The rubric resource gets a "VISA SCORING DISABLED" override prefix the chat reads
+
+Other features are unaffected. If you're a US citizen, a non-US user, or anyone scoring
+roles where sponsorship is a non-issue — turn it off; the rest of the system works.
+
+### Non-US markets
+
+`portals.yml` ships with example shapes for Greenhouse / Ashby / Lever / Workday / Amazon
+/ Google / generic Playwright. Drop in the boards relevant to your market. `modes/rubric.md`
++ `modes/negotiation_playbook.md` + `config/profile.yml` are all yours to localize
+(language, comp ranges, geography priors).
+
+---
+
+## Environment variables
+
+| Var | Default | Purpose |
+|---|---|---|
+| `MCP_JSA_PORT` | `7891` | HTTP port (MCP + file server + tracker dashboard) |
+| `MCP_JSA_HOST` | `127.0.0.1` | Bind host |
+| `MCP_JSA_PROJECT_ROOT` | cwd | Where `cv.md` / `config/profile.yml` / `portals.yml` live |
+| `MCP_JSA_DATA_DIR` | `<install>/data` | SQLite + WAL location |
+| `MCP_JSA_OUTPUT_DIR` | `<install>/output` | Rendered artifacts (PDFs, report HTML) |
+| `MCP_JSA_VISA_SCORING` | `true` | Set `false` to drop visa surface entirely (see above) |
+| `MCP_JSA_LLM_PROVIDER` | `gemini` | Used only by `api`/batch paths: `gemini`, `deepseek`, `none` |
+| `MCP_JSA_LLM_MODEL` | _empty_ | Provider-specific model id |
+| `GEMINI_API_KEY` / `DEEPSEEK_API_KEY` | _empty_ | Provider credentials |
+| `MCP_JSA_SCHEDULER_ENABLED` | `false` | Whether opt-in cron runs at all |
+
+A working starter is at `.env.example`.
+
+---
+
+## Wiring it to Claude Desktop
+
+`npx job_ops-mcp connect` prints the exact config block. The short version:
+
+```jsonc
+{
+  "mcpServers": {
+    "job_ops-mcp": {
+      "command": "npx",
+      "args": ["-y", "job_ops-mcp", "start"],
+      "env": {
+        "MCP_JSA_PORT": "7891",
+        "MCP_JSA_PROJECT_ROOT": "/absolute/path/to/your/job-search/dir"
+      }
+    }
+  }
+}
+```
+
+(Drop into `~/Library/Application Support/Claude/claude_desktop_config.json` on macOS,
+`%APPDATA%/Claude/claude_desktop_config.json` on Windows. Restart Claude Desktop.)
+
+Generic MCP clients that take a streamable-HTTP URL: point them at
+`http://127.0.0.1:7891/mcp` after `npx job_ops-mcp start` is running.
+
+### LibreChat
+
+`npx job_ops-mcp connect` also prints a `librechat.yaml` block. Two shapes:
+
+- **LibreChat as a host process:** `type: streamable-http`, `url: http://127.0.0.1:7891/mcp`.
+- **LibreChat in Docker:** swap to `http://host.docker.internal:7891/mcp` AND allowlist
+  the address under `mcpSettings.allowedAddresses` (LibreChat blocks private/internal
+  addresses by default as SSRF protection). On Linux, also add
+  `extra_hosts: ["host.docker.internal:host-gateway"]` to the LibreChat service in your
+  `docker-compose.yml`.
+
+(Refs:
+[librechat.ai/docs/.../mcp_servers](https://www.librechat.ai/docs/configuration/librechat_yaml/object_structure/mcp_servers),
+[features/mcp](https://www.librechat.ai/docs/features/mcp).)
+
+---
+
+## Working `evaluate_job` payloads
+
+### Step 1 — paste a JD or URL
+
+```jsonc
+{
+  "method": "tools/call",
+  "params": {
+    "name": "evaluate_job",
+    "arguments": {
+      "input": "https://jobs.ashbyhq.com/example/123",
+      "mode": "chat",
+      "title":   "Builder PM",
+      "company": "Frontier AI Tools Co"
+    }
+  }
+}
+```
+
+Returns the rubric, the report format, the active career packet, and a `job_id`. The
+chat client uses those to score + draft the 6 blocks.
+
+### Step 2 — finalize
+
+```jsonc
+{
+  "method": "tools/call",
+  "params": {
+    "name": "evaluate_job",
+    "arguments": {
+      "job_id": "<from step 1>",
+      "mode": "chat",
+      "report": {
+        "archetype_detected": "Agentic / LLMOps hybrid",
+        "block_role_summary": "…",
+        "block_cv_match":     "…",
+        "block_level":        "…",
+        "block_comp":         "…",
+        "block_personalize":  "…",
+        "block_interview":    "…",
+        "block_legitimacy":   "…",
+        "keywords":           ["builder pm", "agentic workflows", "…"]
+      },
+      "scores": {
+        "resume_fit": 86, "taste_fit": 92, "visa_fit": 88, "score_total": 88,
+        "reasoning": "Strong match on agentic workflows + PRDs + SQL/Python.",
+        "concerns":  "Evals experience is adjacent rather than LLM-eval-specific.",
+        "role_category": "pm",
+        "seniority":     "senior"
+      }
+    }
+  }
+}
+```
+
+Server persists, renders HTML at `/files/reports/<id>.html`, returns the URL.
+
+---
+
+## Advanced / outreach features (optional)
+
+### Importing your LinkedIn network → warm-intro finder
+
+Download your LinkedIn data export (Settings → Data Privacy → Get a copy of your data →
+Connections), then:
+
+```bash
+# Through your MCP chat:
+import_linkedin path="/absolute/path/to/Connections.csv"
+```
+
+Now `find_warm_intros(company="…")` returns the people you actually know who work there
+(filtered to non-recruiters, sorted by engineering / leadership weight).
+
+### Importing DOL OFLC H1B data → visa-friendliness signal
+
+Download a quarterly LCA disclosure CSV from <https://www.dol.gov/agencies/eta/foreign-labor/performance>,
+then:
+
+```bash
+# Through your MCP chat:
+import_h1b path="/absolute/path/to/LCA_Disclosure_Data_FY2025_Q1.csv"
+```
+
+`visa_signal(company="…")` then returns a friendliness band (`strong | mixed | weak |
+none`) computed from filings count + recency. **Internal only** — never surfaced in any
+resume, cover letter, or outreach (see the visa hard rule).
+
+If you disabled visa scoring (`MCP_JSA_VISA_SCORING=false`), these tools don't appear in
+`tools/list` at all.
+
+### Scheduler (opt-in cron)
+
+Off by default. To run scans + batch rates on a schedule:
+
+```bash
+# In your MCP chat:
+scheduler_enable jobs=["scan_portals_4h", "batch_evaluate_30m", "daily_digest_morning"]
+```
+
+Job cadence is fixed (4h / 30m / hourly with an 8AM digest window). Toggle off with
+`scheduler_disable`. Survives only as long as the server process is alive.
+
+---
+
+## Hard rules baked in
+
+1. **Never surface visa / work-auth** in any resume, cover letter, or outreach. Visa data
+   is internal scoring only.
+2. **Never invent claims** not in `career_packet`. The materials generator validates LLM
+   output against the packet before persisting.
+3. **Human-in-the-loop everywhere.** No tool auto-submits an application or auto-sends a
+   DM. `apply_prefill` is preview-only — it opens the form in Chromium, drafts values,
+   takes a screenshot, and stops. You submit manually.
+4. **Strict-JSON parsing on the api path** with a recorded `PARSE_ERROR` fallback —
+   never silent zeros.
+5. **Tracker / application / outreach writes are serialized** behind a single write lock.
+
+---
+
+## Layout
+
+```
+job_ops-mcp/
+├── modes/                     # MCP resources (edit me to tune the brain)
+├── templates/                 # CV HTML/LaTeX templates + cover-letter template
+├── fonts/                     # Space Grotesk, DM Sans (woff2 subsets)
+├── cv.example.md              # → cv.md after init
+├── config/profile.example.yml # → config/profile.yml after init
+├── portals.example.yml        # → portals.yml after init
+├── src/                       # TypeScript source (not published)
+│   ├── cli.ts                 # init / start / doctor / connect
+│   ├── server.ts              # HTTP + MCP boot
+│   ├── core/                  # llm, providers, jobs, reports, render, scan_engine, …
+│   ├── http/                  # express app + dashboard
+│   ├── mcp/                   # define + register + tools/
+│   └── migrations/*.sql       # SQLite migrations
+└── data/, output/             # gitignored runtime state
+```
+
+---
+
+## Attribution
+
+- The HTML CV template, font set, and ATS unicode-normalization logic are ported from
+  [santifer/career-ops](https://github.com/santifer/career-ops) (MIT licensed). The
+  6-block A–F report shape, scoring rubric framing, and outreach tone rules are also
+  inspired by that project. **Not affiliated with or endorsed by career-ops** — this is
+  an independent fork of those publicly-released ideas into the MCP transport.
+- The 3-dimension scoring formula (resume / taste / visa), the schema shape
+  (companies / jobs / outreach / enrichment / career_packet views), and the strict-JSON
+  rater rubric are distilled from a personal pipeline ("JSA") that predates this
+  project.
+
+---
+
+## Releasing (maintainer notes)
+
+Releases ship to npm via the GitHub Actions workflow at
+[`.github/workflows/publish.yml`](.github/workflows/publish.yml). The workflow fires
+**only on pushing a version tag** (`v*.*.*`) — never on a push to `main` — so merging
+work never auto-publishes.
+
+### One-time setup
+
+1. **Generate an npm automation token** at [npmjs.com](https://www.npmjs.com/) →
+   click your avatar → **Access Tokens** → **Generate New Token** → choose
+   **"Automation"** (NOT *Read-Only* and NOT *Publish*; Automation tokens bypass 2FA,
+   which CI needs).
+2. **Add it to GitHub.** In the repo → **Settings** → **Secrets and variables** →
+   **Actions** → **New repository secret** → name **`NPM_TOKEN`**, value the token
+   you just copied (starts with `npm_`).
+
+### Cutting a release
+
+```bash
+# 1. Bump the version in package.json. Either edit by hand, or:
+npm version patch      # 0.3.0 → 0.3.1 (also creates a git commit + tag)
+# npm version minor    # 0.3.0 → 0.4.0
+# npm version major    # 0.3.0 → 1.0.0
+
+# 2. If you edited package.json by hand instead of `npm version`, commit it:
+# git add package.json && git commit -m "release: vX.Y.Z"
+# git tag vX.Y.Z
+
+# 3. Push the commit + tag.
+git push && git push origin vX.Y.Z
+```
+
+That tag push triggers `publish.yml`, which:
+
+1. Checks out the tagged commit.
+2. Sets up Node 20 with the npm registry.
+3. Verifies the tag (`vX.Y.Z`) matches `package.json`'s `version` — fails fast on a typo.
+4. `npm ci` + `npm run build`.
+5. `npm publish --access public --provenance` — provenance attaches a sigstore
+   attestation visible on npmjs.com showing exactly which GitHub Actions run produced
+   the tarball.
+
+Watch progress in the repo's **Actions** tab. On success the new version appears on
+[npmjs.com/package/job_ops-mcp](https://www.npmjs.com/package/job_ops-mcp).
+
+## Contributing / feedback
+
+Issues + PRs welcome. There's no contributor guide yet — open an issue first if you're
+planning a large change.
+
+---
+
+MIT — see [LICENSE](./LICENSE).

package/config/profile.example.yml

@@ -0,0 +1,67 @@
+# job_ops-mcp profile configuration.
+#
+# Copy to `config/profile.yml` (or let `npx job_ops-mcp init` do it) and fill in your
+# details. This is the source of truth for your identity + targeting + comp/location
+# preferences. The chat reads it via the rubric + career_packet resources.
+
+candidate:
+  full_name:     "<Your Full Name>"
+  email:         "<you@example.com>"
+  phone:         "<+country area-number>"
+  location:      "<City, State / Country>"
+  linkedin:      "linkedin.com/in/<your-handle>"
+  portfolio_url: "https://<your-portfolio>"
+  github:        "github.com/<your-handle>"
+  twitter:       ""    # optional
+
+target_roles:
+  # Your North-Star roles — what you're optimizing for.
+  primary:
+    - "<role title 1>"
+    - "<role title 2>"
+  # Archetypes drive scoring (`role_category`) + tailoring per JD.
+  # fit values: primary | secondary | adjacent
+  archetypes:
+    - { name: "<archetype label, e.g. AI/ML Engineer>", level: "Senior",      fit: "primary"   }
+    - { name: "<archetype label, e.g. Technical PM>",   level: "Senior/Staff",fit: "secondary" }
+    - { name: "<archetype label, e.g. Forward-Deployed>",level: "Mid-Senior", fit: "adjacent"  }
+
+narrative:
+  # One-line professional headline.
+  headline:   "<your one-line headline>"
+  # Two-sentence story about what makes you uniquely hireable.
+  exit_story: "<your story>"
+  # Top 3–5 differentiators (used in score reasoning + outreach hooks).
+  superpowers:
+    - "<superpower 1>"
+    - "<superpower 2>"
+    - "<superpower 3>"
+  # The 2–4 things that excite you (used in taste_fit scoring).
+  likes:
+    - "<like 1, e.g. 'lean teams, fast iteration, AI/data tooling'>"
+    - "<like 2>"
+  # The dislikes that should drop taste_fit below 50.
+  dislikes:
+    - "<dislike 1, e.g. 'legacy Java enterprise modernization'>"
+    - "<dislike 2>"
+  # Proof points — projects/articles/case studies with measurable impact.
+  proof_points:
+    - { name: "<project name>", url: "<url>", hero_metric: "<short hero-metric phrase>" }
+    - { name: "<project name>", url: "<url>", hero_metric: "<...>" }
+
+compensation:
+  target_range:         "<currency-min – currency-max>"   # e.g. "$130K-180K"
+  currency:             "USD"
+  minimum:              "<walk-away number>"
+  location_flexibility: "<e.g. 'Remote preferred; 1 week/month on-site possible'>"
+
+location:
+  country:              "<country>"
+  city:                 "<city>"
+  timezone:             "<TZ abbr, e.g. PST>"
+  visa_status:          "<e.g. 'No sponsorship needed' or 'On OPT, need H1B sponsorship'>"
+  # Optional. If your target locations differ from `city`, list them here.
+  onsite_availability:  ""
+
+cv:
+  output_format: "html"   # "html" (default) or "latex"

package/cv.example.md

@@ -0,0 +1,53 @@
+# CV — <Your Full Name>
+**Location:** <City, State / Country> (open to <relocation policy>)
+**Email:** <you@example.com>
+**Phone:** <+country area-number>
+**LinkedIn:** linkedin.com/in/<your-handle>
+**GitHub:** github.com/<your-handle>
+**Portfolio:** <your-portfolio-url>
+
+## Professional Summary
+<2–3 sentences. State your specialism, the size/shape of the work you've owned, and the
+hybrid skillset that makes you hire-able for the roles you actually want. Keep it
+concrete — name systems, scopes, and one or two outcomes.>
+
+## Work Experience
+
+### <Company> — <Role Title>
+<Location / Remote> · <Start Mon YYYY> – <End Mon YYYY or Present>
+Products owned: <comma-separated list, optional>.
+- <One-sentence bullet, action verb start, ~15–25 words, real metric inline.>
+- <Second bullet — different verb, different shape of impact.>
+- <Third bullet — emphasise scope (people / systems / dollars).>
+- <Fourth bullet — emphasise a hard-to-mimic technical/product skill.>
+- <Fifth bullet — emphasise a cross-functional or leadership thread.>
+
+### <Earlier Company> — <Role>
+<Location> · <Mon YYYY> – <Mon YYYY>
+- <Bullet.>
+- <Bullet.>
+- <Bullet.>
+
+### <Earliest Company> — <Role>
+<Location> · <Mon YYYY> – <Mon YYYY>
+- <Bullet.>
+- <Bullet.>
+
+## Projects & Open Source
+
+- **<Project name>** (Open source / Personal / Academic) — <one-sentence description with
+  the credibility marker; stack named explicitly>. <github.com/you/repo or link>
+- **<Project name>** — <description>. <link if public>
+- **<Project name>** — <description>.
+
+## Education
+- **<Degree>**, <Institution> — <year range>. <optional 1-line coursework / honors>
+- **<Earlier degree>**, <Institution> — <year range>.
+
+## Skills
+- **AI / LLM Systems:** <list>
+- **Data & Analytics Engineering:** <list>
+- **Infrastructure & DevOps:** <list>
+- **Product:** <list>
+- **Web & Tools:** <list>
+- **Languages:** <primary first>