Taylor Sando
I make AI-agent systems reliable enough to ship, and I write about how software actually gets made.
Psychology and computer science at the University of Manitoba, HCI research published at ACM APGV and IEEE PacificVis, then a decade of product engineering. Now I run a supervised multi-agent development fleet, write about agents, comprehension, and keeping systems legible, and consult with teams putting AI agents into real products.
AI agents, orchestration, and internal automation attached to real product and workflow surfaces.
Reviews, risk maps, decision records, and scoped implementation around reliability, auditability, routing, and system boundaries.
Control-plane tooling across Mesh, Orchestrator, OCC, and Console, the active Sitelayer product loop, product engineering at ReLease/Cios, and earlier HCI research at ACM APGV and IEEE PacificVis.
Beyond consulting I run a few independent ventures, a neighbourhood buying co-op, a live-work building, and a research lab, under one studio. They live on the lab: the Sando studio →
The public version of the work is narrower than the private stack behind it. This site records the systems, research, writing, and professional practice without functioning as an intake funnel.
Practice Areas
AI agent delivery systems
Move from ad hoc prompts and toolchains to governed workflows with clear boundaries, approvals, and operator visibility.
Control-plane reliability
Untangle routing, auditability, review hooks, knowledge surfaces, and failure paths before complexity hardens into production risk.
Ontology and semantic design
Use control-plane, digital, and payment ontology work when delegation, evidence, or system vocabulary needs to stay coherent across teams and tools.
The research and product threads lower on the page show how the same concerns recur across AI-assisted systems: supervision, evidence, routing, and a legible control plane.
Consulting Practice
AI Agent Readiness Review
For teams adopting coding agents or workflow automation but lacking clear boundaries. I map the failure surfaces, approval paths, and missing control-plane pieces before complexity hardens.
Deliverable: written review, risk map, and prioritized next steps.
Control-Plane Audit
For teams that already have agents, queues, or orchestration in play and need clearer routing, auditability, and reliability. I focus on the seams where supervision, policy, and operator control start to fail.
Deliverable: architecture readout, gap analysis, and implementation sequence. Scoped to the system.
Embedded Advisory
For founders and engineering leads who want ongoing help shaping architecture, agent policy, and execution decisions while the system is still moving.
Deliverable: recurring decision support, reviews, and implementation guidance while the system is moving.
The Consulting Practice page records the scope, common artifacts, and boundaries in more detail.
Writing
Experience
Publications
- Effects of animation, user-controlled interactions, and multiple static views in understanding 3D structures (2009) — ACM APGV — with Melanie Tory and Pourang Irani
- Impact of group size on spatial structure understanding tasks (2011) — IEEE PacificVis — with Melanie Tory and Pourang Irani
Earlier Research
- Does the stress-axis in lake sturgeon undergo habituation? — undergraduate honours research with Gary Anderson, listed in 2009 CV material as a Prairie Universities Biological Symposium (PUBS 2008) presentation
Sandolab — Applied Research Lab
Tracking & Overlay Systems
Work across sando-tracker, learning-overlay, Learn, and personal-usability: monitor-focus experiments, overlay surfaces, learning workflows, and desktop attention systems.
Control Plane
Current control-plane repos spanning Mesh, Orchestrator, OCC, and Console: file activity tracking, agent routing, policy tooling, and operator surfaces.
Sitelayer
Construction operations platform for blueprint takeoff, estimates, crew scheduling, rentals, inventory, QBO sync, and feedback capture tied back to reviewable work.
Digital Ontology
A living research corpus on delegation, authority, retrieval, trust, and control surfaces across the digital stack, with the thesis kept in sync with Mesh-backed belief maintenance.
WinWar
Strategy-game research around engine semantics, browser projection, replay, inspection, and guided onboarding for complex state.
Hockeypedia
NHL data project focused on trade lineage, player/game ingestion, and analytics surfaces.
Learn
Standalone Mesh-backed personal learning workbench: Today queue, library and PDF reader, spaced recall, glossary, ontology atlas, proof workflows, sessions, and diagnostics.
The maintained account of shipped work is the Build Record. Live systems and experiments are documented at sandolab.xyz.
Research Threads
Background
University of Manitoba background in psychology and computer science, with research work in animal behavior and HCI. Published at ACM APGV and IEEE PacificVis before moving into consulting and product engineering.
The work spans a consulting practice, control-plane tooling, Sitelayer, digital-ontology and payment-ontology notes, capture/product-trace research, and public experiments including sandolab.xyz, winwar.sandolab.xyz, and hockeypedia.org.