Taylor Sando

I help engineering teams navigate AI-augmented development — architecture reviews, agent orchestration, and system design. I can look at a system and tell you where it breaks, what’s missing, and how the pieces should fit together.

Two decades building software, studying mathematics, and researching how agents — human and artificial — form and update models of the world. B.Sc. Psychology (neuroscience and behavioral research), B.Sc. Computer Science (HCI research, peer-reviewed publications), early-stage SkipTheDishes, and years of consulting across SaaS, e-commerce, and startup product development. Currently building infrastructure for auditable, event-driven agent orchestration.

Services

Architecture Reviews

Evaluate how your frontend, API, database, workflow, and infrastructure layers connect. Identify where the seams are, what’s fragile, and how the pieces should fit together. Not “learn React” but “understand why your current stack composition works or doesn’t.”

AI Agent Strategy

How to adopt AI coding agents without losing accountability. Progressive autonomy frameworks, permission policies, audit infrastructure, and multi-agent coordination. Based on a working system I’ve built and operate daily across Claude Code, Gemini CLI, and Codex.

Tech Stack Assessment

Evaluate whether your stack composition is coherent for your stage and scale. Where the boundaries between services should live, which abstractions are earning their keep, and where complexity is hiding. Security review of AI agent tooling and MCP plugin supply chains.

Writing

Experience

Control Plane 2025 – present
Private infrastructure for orchestrating AI coding agents with full audit trails. File-system activity tracker with CRDT-based versioning (Go + SQLite), multi-agent task orchestrator routing work to Claude Code, Gemini CLI, and Codex with progressive autonomy controls, and a declarative configuration engine generating tool-specific settings from org-mode files.
Cios / ReLease 2023 – 2026
Lead Engineer — Rent deposit insurance platform incubated through North Forge Technology Exchange. Architected the frontend monorepo (React/Redux), designed XState-driven state machines for high-stakes claim workflows, and built deep audit/observability layers including the "Evidence Envelope" pattern. Full stack across MongoDB, Firestore, and GCP. Deel payroll integration, Stripe payment processing.
Datomar Labs 2019 – 2023
Software Engineer — Full-stack development across a portfolio of SaaS products and client projects. Built Brownie Points, a gamified reward system for children with NFC cards and chatbot integration. Developed Convai AI, a conversational bot platform with NLP, SIP/Twilio voice integration, and Elasticsearch (Vue, Go, Postgres). Built the DeutzFahr dealer/wholesaler backend (Go/Postgres) and Braise, a marketplace delivery integration service with Kubernetes. Shopify app development, custom inventory management, and multiple client MVPs.
Freelance & Contract 2012 – 2019
Technical Consultant — Eight years of independent consulting across multiple clients and projects. Built Playbook Edge, a business strategy planning application for InVision Edge (Clojure/Datomic, later Node.js). Developed Pray4EveryHome, a faith-based community platform (2016–2018). Built MVPs and prototypes for PTI Corp (US), Exchange Agency, EdgeCorp Group, and Soliton Apps. Co-founded ScheduleAid at Startup Weekend. Worked with D3.js data visualization, Heroku, New Relic. Built a poker bot with machine learning. Active in the Winnipeg startup community through Meetup and AngelList. Self-directed study in mathematics, Clojure, and functional programming throughout.
SkipTheDishes 2014
Early Stage Engineer — Part of the early engineering team during the company's rapid growth phase. Built initial platform features. Mobile deep linking, TechFutures accelerator, Microsoft BizSpark program.
University of Manitoba B.Sc. Computer Science — 2007 – 2013
Pivoted to Computer Science and joined the HCI Lab under Dr. Pourang Irani. Research on spatial cognition, 3D data visualization, and collaborative information systems. Worked across multiple research labs doing programming, statistical analysis, and data processing. Collaborated with researchers at University of Victoria and University of Alberta. Submitted work to CHI and published at ACM APGV and IEEE PacificVis. Continued as lab instructor for psychology courses throughout. Built TheHockeyPedia — an early hockey data project in Python and MySQL.
University of Manitoba B.Sc. Psychology — 2002 – 2007
First Class Honours. History minor. Dean’s Honour List (2005–2008). Hotchkiss Brain Institute National Neuroscience Competition finalist (2007). Research focus on neuroscience and behavioral biology — was pursuing psychoneuroimmunology before pivoting. Worked across several research labs — animal cognition under Joseph Pear, LabView programming, video tracking systems, cortisol assays, and statistical analysis. Lab instructor for psychology courses for four years, including lecture components. Applied for NSERC USRA and explored grad school before pivoting to Computer Science.

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
  • Does the stress-axis in lake sturgeon undergo habituation? (2008) — Prairie Universities Biological Symposium (PUBS) — with Gary Anderson

Sandolab — Applied Research Lab

Augmented Reality

SandoLens

AI overlays projected into your field of vision. Developer HUD shows agent status, build pipelines, and voice commands. Field mode handles construction checklists, part identification, crew translation, and compliance documentation.

Agent Infrastructure

Control Plane

Monorepo for auditable multi-agent orchestration. Mesh (file watcher + CRDT versioning + LLM summaries), Orchestrator (multi-model task routing via git worktrees), OCC (declarative permission engine), Console (web dashboard + API gateway). Progressive autonomy: agents earn trust incrementally.

Sports Analytics

Hockeypedia

The only platform tracking complete trade trees and computing cumulative games-played impact across asset descendants. Real-time ingestion, AI-powered sprite analysis, content-ready outputs for creators and teams.

Voice & Input

Voice Tools

On-device STT/TTS (faster-whisper large-v3 on RTX 3090 + Piper). Sub-second transcription, GPU-resident model, integrated into Emacs and agent workflows. Your voice never leaves your machine.

6 products, 10 research directions, and a full architecture at sandolab.xyz

Research Directions

Epistemic Circles Knowledge as recursive hierarchies using Ologs for path equivalence and Sheaves for global consistency
Chronotopology Drift detection via persistent homology — measuring topological change, not timestamps
Progressive Autonomy Trust Calculus Formal framework for agents earning autonomy incrementally across three trust tiers
Categorical AR Learning Overlays Functors and sheaf cohomology for adaptive tutorial guidance on closed-source software
Belief Drift Prediction Epistemic health scoring for multi-agent systems — predicting failures from model divergence

Background

B.Sc. Psychology (neuroscience, behavioral research, animal cognition), then B.Sc. Computer Science and HCI research — two decades asking the same question in different contexts: how do agents form beliefs about their environment, and how do they coordinate when those beliefs diverge? That question started with fish in a lab, moved through spatial cognition and 3D visualization research, and now applies to AI coding agents with filesystem access.

I study mathematics — abstract algebra, differential geometry, real analysis — because the structures that show up in distributed systems (CRDTs, Merkle trees, session types) are mathematical objects, and understanding them formally changes how I design software. I maintain a living, falsifiable model of my beliefs about how systems work, with confidence scores and a revision protocol. It is the scientific method applied to software engineering.

Read the full story →