| AIGC |
|
|---|
AI Automation Engineer · Python Developer · AI Agent & MCP Builder
I build agentic systems that survive contact with production: tool-using agents with explicit permission gates, evaluation suites, traces and cost controls — then wire them into real stacks via APIs, webhooks and MCP servers.
Based in Lahore, Pakistan · Portfolio: marvis.web.app
- AI agents with governance — tool permissions, approval steps, execution traces, cost ceilings, and eval harnesses that measure accuracy before anything ships.
- MCP servers and integrations — Model Context Protocol servers that expose real business systems (commerce, analytics, search data) to Claude, Cursor, VS Code and Codex.
- Python automation — FastAPI services, scheduled pipelines and CLI tools that replace manual recurring work.
- Technical SEO & marketing automation — crawlers, audit engines and conversion-tracking diagnostics that turn marketing data into deterministic fixes.
| Project | What it does | Stack |
|---|---|---|
| shopify-growth-mcp | MCP server with 61 tools for Shopify store operations: SEO and GEO audits, deterministic analytics, inventory intelligence, feed quality, approval-gated marketing workflows | TypeScript, MCP, Shopify GraphQL |
| ai-agent-testing-lab | Reliability and evaluation harness for AI agents: labelled suites, CI integration, accuracy/calibration/cost/latency reporting | Python, pytest, CI |
| ai-marketing-shared | Shared foundation for agents: Pydantic schemas, provider abstraction, tracing, cost controls, eval harness | Python, Pydantic v2 |
| envscope | Audits .env files against .env.example and a JSON schema without ever printing secrets |
Python CLI |
| ai-search-visibility-auditor | GEO toolkit: validates AI crawler access in robots.txt, generates llms.txt, checks AI-citation readiness |
Python, Streamlit |
| searchpulse-ai | Search Console intelligence: traffic-drop diagnostics, content decay detection, evidence-based SEO decisions | Python, FastAPI |
| seo-doctor | Technical SEO crawler: broken links, canonical validation, crawl comparison, health scoring | Python CLI |
| conversionbridge | Server-side conversion gateway to Meta Conversions API and GA4 with privacy controls | Python, Firebase |
Languages — Python, TypeScript, JavaScript, SQL, PHP AI / agents — Model Context Protocol (MCP), LLM APIs, tool calling, agent evaluation, RAG, prompt engineering Backend — FastAPI, Pydantic v2, Node.js, REST, webhooks, SQLite, PostgreSQL Frontend / data — Streamlit, Astro, Next.js, Tailwind Infrastructure — GitHub Actions, Firebase, cPanel deployments, git-based workflows
- An autonomous GitHub maintenance engine: issue triage, PR quality gates, duplicate detection, release notes — with persistent state so work resumes cleanly after interruption.
- Demand-validated open-source tooling for AI agent operations (checkpointing, observability, permission auditing).
- 50 merged pull requests across 25 upstream projects (verified 2026-11-20) — MCP tooling, Python libraries, agent runtimes and developer CLIs, including
vibgrate/cli,pygrpc/pygrpc-native,supabase-py,trafilaturaandmcp-memory-service. - 60 pull requests open for maintainer review.
- Each contribution ships with a reproducible case or test, and stays inside the project's contribution guidelines.
Demonstrated in shipped repositories — Python development, MCP server development, API integration, agent evaluation harnesses, technical SEO automation, conversion tracking, CLI tooling, CI setup.
Actively learning — distributed agent orchestration patterns, infrastructure-as-code, cloud deployment at scale.
Interest areas — agent security and governance, evaluation-driven development, developer productivity tooling.
I'm open to conversations about AI automation projects, MCP integrations, Python development and technical SEO automation — including open-source collaboration and contract work.
- WhatsApp: +92 316 7417135
- Website: marvis.web.app
- GitHub: @hamids-ai-automations
All projects listed above are public repositories. Status labels (released / beta / in development) are stated in each repository's README; nothing here is presented as a finished product unless it is one. (内容由AI生成,仅供参考)


