Systems Decision Architect · Applied AI Runtimes · Catastrophe Risk & Spatial Systems · Database Internals

I design and build the infrastructure through which complex decisions get made under high uncertainty. My work focuses on deterministic agent runtimes, sandbox isolation, spatial loss accumulation engines, and high-throughput data kernels.

• Agent Runtimes & Execution Security: seccomp-bpf system call filtering, POSIX process boundaries, resource budgeting, MCP orchestration, deterministic execution containers.
• Spatial & Actuarial Modeling: Sub-kilometer wildfire/flood accumulation grids, $25M+ PML exposure modeling, high-speed spatial indexing, catastrophe underwriting infrastructure.
• Decision Infrastructure & Data Internals: Reusable decision engines under uncertainty, event-driven state evaluation, zero-leakage enterprise boundaries, high-performance C++/Python data kernels.
• Distributed Fleet & Agent Orchestration: Heterogeneous multi-node compute coordination, RPC/event messaging buses across cloud and local nodes, deterministic state graphs, and zero-trust boundary governance.

Working on low-level correctness, isolation boundaries, and spatial kernels in the broader ecosystem:

• DuckDB Spatial: Resolving WAL replay non-flat vector appends in RTREE index checkpointing (#864 — verified 20/20 cross-platform CI).
• Anthropic Sandbox Runtime: Hardening container write boundaries (#518) and designing eBPF supervisor monitors for read violation auditing (#511).
• OasisLMF: Automatic CRS reprojection for multi-peril risk accumulation in RTREE spatial joins (#2138).
• Model Context Protocol: POSIX process isolation and resource limits for host-supervised stdio servers (#3457).

• Live Systems Project Board: github.com/users/connerkup/projects/1
• Portfolio: connerkupferberg.com
• GitHub: @connerkup
• LinkedIn: linkedin.com/in/connerkupferberg