I/We/Us, the squad, build and commission products and systems. We hunt for problems to solve.
This site is the drawing set for that work. Each sheet has a title block, a revision, and a status. Where a build is unfinished, the sheet says so.
Two machines and no cloud GPUs. They run open-weight models behind one router, so a multi-agent council can review work without paying per token.
The worker rejected every council-sized packet. The model was not the limit. The serving configuration was giving each request far less context than the model could support. The fix was routing rather than a bigger model: long packets go to the senior node, and short ones stay on the worker.
Check the running configuration before blaming the component. The spec sheet and the running config disagreed, and the running config is what determines behavior.
Five phases with a gate between each one. Nothing moves forward on enthusiasm. OpsG spans every phase, which means the people who will operate the thing get a say before it is built, not after.
A multi-agent market intelligence platform. It combines a geopolitical signal tracker, sector rotation, macro stress indicators, and environmental signal monitoring behind one AI chat interface.
HTML/CSS/JS · Vercel serverless · Clerk auth · Redis (Upstash) · Claude Sonnet, Groq Llama 3.3 70B, Gemini · Resend · Alpaca API
OpsG
A driver-owned peer-to-peer rideshare platform. I designed the go-to-market architecture and the agent-backed matching logic.
The two-sided liquidity problem: for many people, “I need a ride” automatically means opening Uber or Lyft, not because of deep loyalty but because those apps own the behavioral default. Years of repetition have created a simple habit loop
So Bump or any other has to solve two problems at once: create enough concentrated driver supply to make the marketplace work, and interrupt the rider’s existing routine strongly enough to get them to try something else.
OpsG does not replace PLC, DCS or SIS control logic. IEC 61508 and ISA/IEC 61511 place strong lifecycle, verification, and safety-integrity requirements around systems performing safety functions, while NIST emphasizes that OT has unusual safety, reliability, and availability constraints.
ISA-95 already gives us a natural architectural boundary: Levels 0–2 contain the physical process, sensors/actuators, PLCs/DCS and supervisory control; Level 3 handles manufacturing operations; Level 4 handles business planning/ERP.
An illustrative placement of OpsG in a plant could look like this:
Different industries. Same Operative Grammar.
OpsG coordinates complex digital and physical work across the systems organizations already use, helping teams move from operational need to verified result without replacing their existing technology stack.
Illustrative use cases describing the kind of problem OpsG is built for. They are not customer deployments and have not been independently validated.
A large multi-site distributor operates refrigerated and dry distribution centers, manages thousands of SKUs, and coordinates warehouse, transportation, maintenance, and retailer delivery operations.
A refrigerated trailer begins drifting outside its target temperature while several customer deliveries remain on the route. The information needed to respond is scattered across telematics, dispatch, route planning, inventory, delivery windows, and food-safety procedures.
Illustrative response pattern: understand the situation, apply existing governance, coordinate the approved action, and verify the result.
OpsG helps operations teams resolve exceptions across transportation, warehouse, maintenance, and enterprise systems while preserving existing responsibilities and controls.
Existing WMS, TMS, telematics, CMMS, and ERP remain in place. OpsG coordinates work across them without requiring replacement of the underlying systems.
A multi-site industrial manufacturer operates production lines, quality systems, maintenance programs, warehouses, and engineering functions across a large plant network.
A production line begins producing material outside a quality limit. Maintenance, process conditions, recent changes, material lots, quality results, and production schedules all need to be considered before the correct response can be determined.
Illustrative response pattern: understand the situation, apply approved operating limits, coordinate the permitted response, and verify the result.
OpsG helps manufacturing teams coordinate complex responses across production, quality, maintenance, and enterprise systems while respecting existing operating and safety boundaries.
PLCs, DCS, SIS, MES, CMMS, ERP, and quality systems remain authoritative within their existing boundaries. OpsG coordinates work across those systems without requiring their replacement.
A city or county operates across permitting, inspections, public works, constituent services, fleet, facilities, utilities, licensing, procurement, and other public-service functions—often through separate systems, departments, and approval chains.
A resident reports a recurring streetlight outage. The issue touches 311 intake, asset records, GIS, work-order systems, field crews, contractor history, parts availability, and potentially prior unresolved service requests. Today, resolution may depend on staff manually moving between systems and determining what should happen next.
Illustrative response pattern: understand the request, apply departmental responsibilities, coordinate the approved work, and verify the outcome.
OpsG helps municipalities coordinate service work across existing departmental systems while improving accountability, visibility, and operational follow-through.
Existing 311, GIS, permitting, ERP, CMMS, utility, document, and work-order systems remain in place. OpsG coordinates work across them without requiring replacement of the underlying systems.