Staff AI Engineer @ EchoStar (SF Bay Area) — I build the retrieval, agent-orchestration and evaluation pipelines that put LLMs in front of millions of consumers. Before that: agentic data engines on carrier-scale spatial data at Polaris Wireless, APFS test automation at Apple CoreOS, ad-targeting ML at Walmart, ML research at LBNL. UC Berkeley CS '23, 6 published papers.
| Currently | Agents, RAG and evals at consumer-telecom scale · growing the open-source agent-tool family |
| Ask me about | MCP tool servers · sub-agents & agent skills · eval-driven development · prompt-injection defense · vLLM serving · context engineering |
| Reach me | michaelle.lubich@gmail.com · mishalubich.com · LinkedIn |
imsg · imail · inotes · wa · bitbucket-cli · confluence-cli · pdfify-md · jenkins-mcp · twig · like-fable · ical · vgate · own-chrome
Packages: npm pdfify-md · PyPI twig-cli · PyPI mac-ical · PyPI vercel-mcp · PyPI jenkins-mcp-cli · Homebrew tap
Recent builds: voice-agent (offline voice agent, barge-in, p50 643 ms to first audio) · callgen (call recording → HTML readout, multi-agent + fact-check) · claude-tiers (tiered Claude Code delegation plugin) · ai-awesome-chat (drop-in streaming AI chat for Next.js) · scroll-craft (agent skill for scroll-driven sites)
Products: Lupfr (SF music events & talent) · W3Sourcing (recruitment) · EnrichData (AI CRM enrichment) · mishalubich.com
6 peer-reviewed papers in machine learning and hydrology, from UC Berkeley and Lawrence Berkeley National Laboratory → full list on Scholar.
Production agentic AI (MCP, multi-agent, RAG, evals), technical co-founding / advising, or ML research — if that's you, email me or book a call.
