Location: USA / Central Time
Remote: Yes
Willing to relocate: No
Technologies: Agentic engineering systems, Agent Harnesses, AgentOps, multi-agent orchestration, self-improvement loops, LLMs, RAG, evals, Python, FastAPI, MLOps/LLMOps, Kubernetes/GitOps, AWS/GCP/Azure, Terraform, observability, vector databases
Resume: Available upon request
Email: paperaxis.refining795@simplelogin.com
---
Principal, Agentic Engineering Systems · "I ship with agents."
PhD Mathematics · 15+ years production AI/ML.
I build self-improving engineering systems that ship PRs against real enterprise codebases, including personal agents that learn each engineer's context and expertise, and the harness that keeps them in bounds.
Most recently I led an AI tooling team at a Big 4 professional services firm, building production LLM tooling on open-source agent/RAG components. Before that I designed autonomous engineering systems for real enterprise software workflows. They used the same channels any engineer does: repos, PRs, CI, and Slack.
That work, including unsupervised runs on real production code, made the next bottleneck obvious: capability isn't the hard limit anymore. Authority, trust, and blast radius are.
I've hardened the design into an Agent Harness architecture: authority boundaries, blast-radius scoring, multi-agent consensus, and human approval routing. Other teams reached out to learn the approach.
Background: led MLOps adoption across 30+ business units; drove $48M in annual data-cost reduction; built production RAG and Slackbot systems for public sector and Fortune-scale telecom. ML roots in signal processing, frame theory, NLP, time-series, and applied math. US patent pending.
Looking for work building governed engineering-agent fleets: personal agents that learn each engineer's context and expertise, shared team agents that coordinate across repos and workflows, and the control plane that keeps it safe enough to run in prod. Also a good fit for agentic SDLC platforms, internal AI infrastructure, or principal-architect work. Prefer C2C/1099, open to full-time or fractional.
Email for full resume and references.