I'm Jalil Laaraichi — a senior software engineer with 8+ years shipping for enterprises and startups, now focused on LLM agents, tool calling, orchestration, evals, and realtime voice. Creator of Harness Config, an open specification for reviewable AI-agent configuration.
An open specification proposal + TypeScript reference implementation that ends the copy-paste sprawl of AI-tool config files.
Modern repos carry AGENTS.md, CLAUDE.md,
.cursor/, Copilot instructions, and more — the same prompt copied five places,
and only one copy ever changes. Harness Config makes the ownership boundary explicit: durable
config lives in one reviewed source root, then is projected one-way into every live surface
each tool reads.
Local-first developer tools for the agent era — published, documented, and CI-tested.
Repo-local skill switchboard for AI agent skill libraries. Manage, activate, and version agent skills locally — published as @skills-kit/cli on npm.
Private macOS menu-bar app that gives a quiet visual cue when speech is happening around your Mac. On-device Silero VAD — no recording, no transcription, no upload.
Browser QA automation brief authoring: demonstrate interactions in headed Chrome, annotate UI regions, and hand off validated briefs that downstream QA agents drive Playwright-style tests from.
Composable terminal-UI packages for exact-frame TUIs: ANSI helpers, layout primitives, charts, panels, and a typed-routes framework layer with focus zones and transitions.
Enterprise scale and startup speed — eight-plus years across both.
Clients include PagerDuty and Illustrative Mathematics. Architected an agent-orchestration platform on Cloudflare Durable Objects — virtual filesystem, WebSocket sync, tool execution, run tracing, usage budgeting — now in private beta. Delivered a security assessment and analytics pipeline for PagerDuty. Created and maintain Harness Config.
Three concurrent tracks at one of the world's largest agriculture companies: migrating the Digital Hub commerce platform from a legacy web app to a React Native–first application (iOS, Android, web), implementing modern AWS networking and VPC features for the move to federated cloud infrastructure, and shipping data products for the Field Planning app — retail partners designing farmers' season plans (rotation, fertilizer, yield) augmented with satellite imagery and dynamic inventory data.
Led engineering delivery for a project-management SaaS; migrated CI/CD from legacy Jenkins/Ant to GitHub Actions and drove a service-oriented backend refactor with automated browser and integration tests.
Built enterprise portals and integrations for satellite-imagery ordering workflows, connecting customer-facing systems with SAP, CPQ, and Salesforce back ends.
Production TypeScript end to end, with the AI tooling to match.