Make the agent harness measurable.
Mohamed A M Elansary, PhD — production GPT, Claude, and Gemini agent systems with tool use, regression evaluation sets, provenance, and tenant isolation, plus scientific/HPC evaluation and uncertainty quantification.
Production agents and evals
- Ships production GPT, Claude, and Gemini agent workflows with tool use, retrieval, query routing, and production responses at Vertexium.
- Maintains regression evaluation sets and provenance so a tool-using trajectory can be inspected, not just scored.
- Multi-tenant retrieval and data isolation so one tenant's context does not leak into another.
Scientific measurement discipline
- Six-plus years of multimodel, multi-basin forecast experiments that turned hydrologic data streams into controlled evaluations across hydroclimates.
- Uncertainty quantification, imperfect USGS / NOAA / NASA observations, and regime-dependent failure reporting on Linux/HPC.
- That maps to a skill-and-eval contract: intended behavior, honest baselines, and a diagnostic read a non-builder can use.
Proposed first contribution
For one real operational vertical that already produces decisions, define a skill-and-eval contract before expanding the harness. Intended behavior, a small failure taxonomy (overconfident tool use, missing context, skipped check, isolation miss), provenance on traces, simple baselines, and a diagnostic read a non-builder can use, with decision capture live from day one. This is a proposed measurement approach, not a claim of prior Stand product, MCP platform, or fleet-orchestration ownership.
Honest fit boundary
Insurance operations and the Stand product stack are a stretch, as is dedicated agent-platform software-engineering depth such as MCP implementations and a fleet-orchestration substrate. I have not authored insurance or Stand-domain publications, and I do not invent metrics or safety research. The credible contribution is production agentic systems, evaluation and observability, scientific/HPC rigor, and a bias toward the reliable harness rather than one showcase agent.
Role and location
Member of the Technical Staff - Agentic Engineer · San Francisco · Ashby workplaceType Hybrid. The live posting body does not specify a hybrid day count. Willing to relocate to San Francisco with a relocation package. Remote eligibility is not asserted.
Posting compensation: “The annual base salary range for full-time employees in this position is $270,000 to $325,000 + meaningful Equity Grant.” Ashby display “$270K – $325K • Offers Equity”. · Official role posting