Systems. Intelligence. Product. Commercial execution.
Building across distributed systems, AI infrastructure, low-latency software, compilers, agentic automation, and production products — with the commercial depth to connect architecture to customer value, positioning, sales, and execution.
AI-accelerated execution. Human-directed judgment.
Strongest at building useful mental models of unfamiliar systems quickly: isolating the failure mode, execution path, architectural boundary, state transition, test surface, and invariant that actually matters.
AI is used as an engineering and business multiplier — accelerating research, implementation, validation, orchestration, and delivery while keeping problem selection, architectural reasoning, tradeoffs, review interpretation, and final decisions human-directed.
The result is a single execution stack spanning systems engineering, AI, product, and business rather than four disconnected disciplines.
|
Distributed systems, backend architecture, concurrency, memory ownership, low-latency runtimes, compiler/runtime behavior, performance, reliability, and production debugging.
|
Production AI agents, tool-using workflows, orchestration, model/API integration, human-in-the-loop execution, memory, retrieval, evaluation, and automation across OpenClaw, Hermes, LangChain, and custom stacks.
|
|
From architecture and APIs to SaaS products, internal platforms, deployment, observability, billing, analytics, and operational workflows.
|
Product strategy, business development, sales, positioning, branding, marketing, and go-to-market thinking — used to keep technical work connected to adoption and business value.
|
Small, scoped changes built around existing architecture, explicit invariants, regression coverage, and repository-native conventions.
| Project | Engineering focus | Status |
|---|---|---|
| Microsoft TypeSpec | Playground state synchronization; prevented stale Monaco callbacks from overwriting newly loaded sample configuration | Merged · PR #11660 |
| Redpanda | RPC transport memory ownership so backpressure follows queued/send-buffer lifetime rather than timed-out response futures | Open · PR #31594 |
| CockroachDB | Deferred non-transactional index split creation to post-commit backfill to protect aborted schema changes | Open · PR #173411 |
| Supabase Supavisor | PostgreSQL cancellation synchronization across backend reuse | Open · PR #1149 |
| PyTorch | Floating-point signed-zero reduction parity between eager execution and Triton/Inductor | Open · PR #193660 |
| LLVM | X86 lowering for non-power-of-two vector integer division while preserving full-lane vectorization | Open · PR #215076 |
Additional upstream work spans Microsoft Pyright, Meta Pyrefly, Velox, Docker, JAX, Cloudflare Workers, Solana Web3.js, and other production-grade systems.
Open-source contributions are independent upstream engineering work and do not imply employment, partnership, endorsement, or client relationships with the projects or organizations listed above.
Symptom
→ execution behavior
→ architectural invariant
→ root cause
→ minimal durable correction
→ deterministic validation
The goal is not maximum code output. It is the smallest correct change that fits the system.
Founder of Ribbsaeter Systems — a senior engineering studio focused on production software, AI infrastructure, agentic systems, distributed architecture, automation, and direct technical execution.
Technical depth is paired with product and commercial context: understanding not only how a system works, but what should be built, why it matters, how it reaches users, and how to ship it without unnecessary complexity.