Jim Bonant | AI/ML Infrastructure & MLOps Engineer
AI Startup
San Francisco, CAOn-siteFull-time$140k – $220k / year
About this role
Jim Bonant | AI/ML Infrastructure & MLOps Engineer
ABOUT
I build the infrastructurethat makes AI reliable.
Senior AI/ML Infrastructure Architect and MLOps Engineer with 13+ years designing, deploying, and optimizing scalable hybrid cloud and on-prem AI platforms. I specialize in production LLM serving, Retrieval-Augmented Generation (RAG) systems, vector databases, and custom LLM solutions in Python, Rust, and C++.
I’ve led MLOps transformations that delivered 70–80% faster deployments, 99.99% uptime, and substantial cost and performance gains across enterprise environments.
Currently running my own expermintal platform Naturally Artificial of Pennsylvania, I architect bare-metal and containerized LLM inference platforms and containerized AI products that power real customer workloads at scale. Previously I led machine learning teams at PNC and drove major containerization and CI/CD initiatives at BNY Mellon.
GitHub ML OPS↗•
13+
YEARS IN PRODUCTION AI & DEVOPS
99.99%
UPTIME ON CRITICAL PLATFORMS
4×
DEPLOYMENT VELOCITY IMPROVEMENT
100k+
DAILY INFERENCES SUPPORTED
CAREER
Experience
March 2026 — Present
Naturally Artificial of Pennsylvania
Senior AI/ML SecOps & Platform Engineer
Architecting and deploying scalable hybrid bare-metal LLM inference platforms integrated with containerized applications and APIs. Supporting 100k+ daily inferences with sub-500ms latency (3× speedup).
- •Designed and led end-to-end MLOps + DevOps CI/CD pipelines (Jenkins, GitLab) for full-lifecycle model deployment and monitoring.
- •Built custom RAG systems including the S.A.R.A. executive AI agent and domain-specific variants achieving 95% retrieval accuracy.
- •Engineered complete infrastructure stack with comprehensive AI/ML SecOps (zero-trust, vulnerability scanning, encryption, AI-powered security gates).
- •Currently operating live multi-cloud environments on AWS, Azure, and Google Cloud.
- •Deploying highest-availability infrastructure via vLLM, KServe, and G