Skip to content
View benikigai's full-sized avatar
🕹️
Side quest’n
🕹️
Side quest’n

Sponsoring

@openclaw

Block or report benikigai

Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
benikigai/README.md

Benjamin Shyong

San Francisco | Staff TPM at Meta, ex-Apple | Operator-builder for agentic infrastructure

TypeScript Python Next.js PostgreSQL OpenClaw MCP/A2A Codex Claude Tailscale

I build AI systems for messy physical-world operations: factories, guest ops, data pipelines, agent memory, and owner-led businesses where the software has to survive contact with real work.

I spent a decade shipping consumer hardware from prototype to mass production at Apple and Meta. Now I build agentic infrastructure against live operating surfaces: a five-property short-term-rental business, a self-hosted property management stack, agent discovery protocols, and a home inference fleet.

Currently building

  • 🛒 Injester - agent-native commerce for Shopify: catalog to checkout, attribution, and recovery.
  • 🦾 Elias - my OpenClaw assistant: seven agents, five channels, versioned memory, local inference.
  • 🧰 TurnkeyHost - self-hosted PMS for five STRs: reservations, guest comms, cleaning, financials.
  • 🧭 Agent Web Protocol - open discovery spec for what agents can do on a web surface.
  • 🏡 Stay Ikigai - direct-booking STR brand and live testbed for my ops stack.

Hardware and operations

I am a systems and industrial engineer by training. Purdue SIE. Started in supply chain and manufacturing, moved into hardware program management, then Apple and Meta consumer hardware.

At Meta, I work on manufacturing readiness for Ray-Ban Display AR glasses. Before that I spent six years at Apple across iPad Pro NPI, softgoods, and special project manufacturing bring-up. Before Apple, I worked across supply chain, ERP, and factory systems at Lincoln Electric.

The through-line is operational reality: mechanical, electrical, firmware, supply chain, factory ops, yield, throughput, people, and deadlines all in the same room.

A few receipts

  • 1st place, Nebius.Build SF: Injester - agent-commerce prototype that became the Shopify agent-buyable infrastructure.
  • 1st place, DataHub x Nebius: Data-OnCall - 4-agent data-quality incident response with lineage tracing and writeback.
  • 1st place, Cerebral Valley x Rosewood Hospitality 2030: Red Thread - luxury-hospitality agent OS with live A2A handoff.
  • Top-six finalist, AWS Builder Loft Agent Hack Day: LoopOS - multiplayer MCP-native back-office agent for STR operations.

Pinned Loading

  1. injester-ai/Injester injester-ai/Injester Public

    Python 2 1

  2. nightshift nightshift Public

    Controlled agent harness for lights-out software work: persistent Ralph Loops, evaluator backpressure, guardrails, metrics, and git audit trails.

    TypeScript 14 5

  3. elias elias Public

    autonomous ai assistant running on openclaw. 7 agents, 5 channels, persistent memory, home inference cluster.

    1

  4. marginalia marginalia Public

    On-device voice agent for smart glasses. Gemma 4 + Cactus + Even Realities G2. Zero cloud.

    C 3

  5. datahub-hackathon-oncall datahub-hackathon-oncall Public

    4-agent incident response team for data quality bugs — DataHub × Nebius hackathon, April 10 2026 @ EF SF

    Python 2

  6. digitalnomd/altiair digitalnomd/altiair Public

    Resilient edge intelligence for contested and disconnected environments

    TypeScript 3