Benjamin Shyong
Staff TPM at Meta ex-Apple
Consumer hardware by day. Agentic infrastructure by night
Hardware Operations
Consumer AR Glasses Manufacturing
Staff Technical Program Manager for Meta's Ray-Ban Display AR glasses. Led design, engineering, supply chain, and factory ops from prototype through mass production for a first-generation consumer AR product.
Operations Program Manager & Product Quality
Operations Program Manager for iPad Pro, prototype to mass production. Apple Special Projects to stand up a new manufacturer and automation line for Health Tech, then ran ops and quality for iPhone and AirTag.
Manufacturing, Supply Chain & Factory Operations
Manufacturing Engineer to Global Supply Manager to Operations Team Lead in Torreon, Mexico, then an APAC strategic assignment in Shanghai directing an ERP rollout across five China plants.
Operator
- 10 years Consumer Hardware NPI at Apple + Meta
- 9 factory cities: China, Vietnam, Taiwan
- xFN convergence: ME, EE, SW, supply chain, ops
Builder
- 3× hackathon 1st place + 1× top 6 finalist
- Full-stack AI builds: A2A, MCP, LoRA fine-tunes, vLLM
- Five-property STR brand on production software I write
Builds
all projects →
The missing layer between the human web and AI agents. Strips a page to what an agent actually needs. 81% less content, 3 to 5x better task completion.
Multiplayer MCP-native agent that replaces the back office for short-term rental operations. Multiple AI clients share one durable backend.
Two-agent voice architecture for real-time standup facilitation. Scriber runs the meeting by voice and mutates the Linear board live; Mnemo silently remembers across standups and whispers guidance.
Multi-agent content engine for technical founders. It researches what matters, writes in your voice, and a second agent verifies it didn't change a fact to sound smoother.
Manufacturing execution system for autonomous agents. OpenAI's 7-stage AI-Native Engineering Workflow as a bipartite state machine, agents do throughput work, humans own the judgment gates.
A home inference fleet, roughly 1.68TB of unified memory. Two Mac Studio M3 Ultra (512GB each), an M4 Max agent box, a 4-node DGX Spark GPU cluster on a 200Gbps ring, a Mac mini head node, NAS, and a router. Local inference, no cloud.
Hackathon wins
all hackathons →Red Thread
first placeFirst place. Built Red Thread, a hospitality concierge concept rooted in Rosewood's "Sense of Place" thesis.
LoopOS
top 6 finalistTop 6 finalist. Built LoopOS, a multiplayer MCP-native agent that replaces the back office for short-term rental operations.
Data-OnCall
first placeFirst place. 4-agent incident response team for data quality bugs, fine-tuned Llama 3.1 + LoRA for NL→GraphQL, end-to-end in 54.8s at ~2¢/run.
Injester
first placeFirst place. Built the earliest Injester prototype, agent-to-merchant commerce infrastructure. Demoed at NVIDIA GTC the following Tuesday.
About
One of my favorite moments in life is meeting someone chasing the purpose they were made for. Someone living so deep inside their ikigai. That rare alignment of love, mastery, contribution, and craft. You feel it within seconds. The air around them sings.