A catalog of selected works

Selected work.

Six entries spanning eight years, in three media: software, authored frameworks, and adult-learning programs at scale. Each a different scale of the same underlying question.

2018 to Present Vancouver, BC

A catalog rather than a CV. The point is not the timeline of employment but the through-line of practice: how a question first noticed in a single classroom worked its way into organizational learning ecosystems, into AI architecture, and finally into commercial work.

The plates
  1. i. Headwater 2025 →
  2. ii. Insight Engine 2024–2025
  3. iii. Cultivating the Future 2023
  4. iv. UBC Imagine Day 2022–2023
  5. v. Jump Start Leader Training 2022–2023
  6. vi. K-12 distance learning 2018–2021
Plate I

Headwater.

MediumProductized intelligence service

YearOctober 2025 → present

RoleFounder & lead analyst

Problem

Audience-led businesses operate blind to their conversation layer. Generic sentiment tools collapse meaningful signal into “positive / negative / neutral” and miss the dormant advocates, the cooled warm leads, and the ranked product-demand language sitting in the comments.

Context

Educators and enterprise clients with content-driven business models, where audience response is a leading indicator and decisions about format, pricing, and product direction depend on reading the room accurately. Two tracks: a Creator Program for educators and online creators, and a Brands & Studios track for SaaS, gaming, media, and market research.

Built

A productized intelligence service grounded in three meta-analyses encompassing 1,532+ effect sizes and designed against the documented failure modes of commercial sentiment tools. Complete-population analysis at the individual commenter level: every statistic traces to a computed value, every quote traces to a specific person on a specific piece of content.

Sample finding panel
Product demand signal strong · n=47

Viewers explicitly requested deeper content on a specific topic. Ranked first by purchase-intent language.

“I would 100% pay for a full course on this” · viewer comment, 340 likes

Warm leads who went cold 200 viewers

Previously engaged viewers stopped commenting after a format shift. Last comments were positive. They didn’t leave unhappy.

Testimonial inventory 23 found

Comments at 500+ likes expressing outcomes and trust: better sales copy than a copywriter would produce.

Every finding verified against source data.
Outcome

A representative engagement reads 30,000+ comments across 17,000+ unique participants, surfacing 200 dormant advocates and a ranked product-demand inventory. Findings are delivered with citation-grade traceability rather than aggregate scores.

What this proves

I can productize a research methodology into a delivered service: complete-population community analysis with citation-grounded findings, sold to creators and enterprise.

Visit headwater.cc  →
Plate II

Insight Engine.

MediumApplied R&D · software architecture

Year2024–2025

RoleFounder & principal architect

Problem

Generic AI defaults to the wisdom of crowds. For organizations whose competitive advantage is their specific, hard-won knowledge, that’s a liability: the model returns what the internet thinks, not what the organization actually knows. Traditional search isn’t an alternative either, because it requires knowing what to look for.

Context

Twenty-one months of self-funded R&D, motivated by direct experience operating as a “human RAG” through 500+ scattered documents during organizational transition. Tested across three corpora with very different shapes: financial due-diligence packets, academic research synthesis, and property-management regulatory compliance documentation.

Built

A production-oriented KG-RAG prototype: hybrid retrieval over Neo4j and pgvector, automatic knowledge-graph extraction with Schema-on-Edge, citation tracking preserved across multi-hop reasoning, tiered LLM routing for roughly 60% reduction in API spend, and an evaluation harness against established public multi-hop QA benchmarks.

Outcome

The same architecture (same ingestion, same Schema-on-Edge graph, same citation discipline) later powered Headwater’s community-intelligence pipeline. Different corpus, identical engineering. The cross-domain transfer was the validation.

What this proves

I can architect production AI systems end to end: hybrid retrieval, knowledge-graph extraction, citation tracking, and the engineering judgment to keep complexity in service of real organizational use.

The technical deep dive  →
Plate III

Cultivating the Future.

MediumAuthored framework · 12-chapter white paper

Year2023

RoleAuthor · UBC Learning Strategist

Problem

Most institutional leadership programs default to content delivery. Lectures, seminars, modules. The research is unambiguous that this isn’t where development actually happens, but the operational pull toward replicable curriculum is strong, and most programs end up reaching only the students already in leadership roles.

Context

Authored during my tenure as Learning Strategist at UBC’s undergraduate-development programs, and benchmarked against fifteen peer institutions across North America to test where the framework agreed with prevailing practice and where it diverged.

Built

A 12-chapter framework arguing that leadership is emergent, not taught. Organized around five pillars (service orientation, community and belonging, global perspectives, introspection paired with situational awareness, adaptability), grounded in the Social Change Model and the 70-20-10 framework, and explicit about the conditions under which content delivery moves the needle and the ones in which it doesn’t.

Outcome

The position the white paper takes (that development happens through community, belonging, and experience rather than content transmission) later became the philosophical foundation for everything else: the AI adoption work, the Insight Engine’s design philosophy, the Headwater methodology.

What this proves

I can author at the framework level: a 12-chapter leadership-development theory benchmarked against fifteen peer institutions and grounded in adult-learning research.

Read the white paper  →
Plate IV

UBC Imagine Day.

MediumProject sponsorship · operational leadership

Year2022–2023

RoleProject sponsor

Problem

A flagship orientation event reaching 10,000+ incoming first-year students sits on a fragile substrate: critical operational knowledge scattered across hundreds of documents, no systematic way to retrieve it, and continuity dependent on the people who happen to remember.

Context

Project sponsor for UBC’s annual Imagine Day, the university’s flagship undergraduate orientation event. Budget allocation, vendor contracts across 400+ partner organizations, facilities coordination, and the internal knowledge systems that kept programs running.

Built

An operational backbone holding budget, contracts, and program continuity together while documentation lagged behind decisions. Less a designed system than a real-world demonstration of why systematic institutional memory matters when it isn’t there.

Outcome

Program continuity maintained for tens of thousands of participants. The personal cost was operating as a human knowledge graph: months spent manually retrieving answers from 500+ scattered documents to keep programs running. That experience is what motivated the Insight Engine.

What this proves

I can hold operational responsibility at scale, with first-hand evidence of why systematic institutional memory matters when it isn’t designed for.

Plate V

Jump Start Leader Training.

MediumAdult-learning program design

Year2022–2023

RoleLead designer & facilitator

Scale330 staff · 91% satisfaction

Problem

Large staff-training programs face the standard tension between reach and depth. Front-loaded curriculum scales but doesn’t transfer. Hands-on coaching transfers but doesn’t scale.

Context

Designed and led the Jump Start Leader Training program for student-staff at UBC’s undergraduate-development programs. A cohort of 330 to be onboarded with limited in-person time and a high standard for quality.

Built

Communities of practice as the primary development vehicle rather than curricula. Multi-layered mentorship structures (formal mentors, peer coaches, accountability partners). Experiential workshops grounded in 70-20-10. Procedural content externalized into well-organized reference so face-to-face time could go to application, coaching, and judgment.

Outcome

60% reduction in face-to-face training time. 91% participant satisfaction across the cohort of 330. The result that mattered more was the proof that the principle could survive scaling: when the externalization is rigorous and the human-delivery layer is preserved deliberately, programs at scale don’t have to choose between reach and depth.

What this proves

The design principle survives scaling. Externalize procedural content; reserve human capacity for what nothing else can do. The Insight Engine is the same idea, expressed in software.

Plate VI

K-12 distance learning.

MediumAsynchronous curriculum · STEM

Year2018–2021

RoleProgram developer & instructor

ScopeThree BC districts

Problem

K-12 distance learning at the time was largely synchronous lectures pasted into an LMS. The format ignored what asynchronous self-regulation actually requires: well-structured materials, predictable feedback loops, and human attention reserved for the moments where it could not be replaced.

Context

Three BC districts, 150–300+ concurrent students annually, asynchronous self-regulated STEM courses. The COVID-19 shift later added 30+ classroom teachers who needed mentorship transitioning to a delivery model their pedagogical training hadn’t, in most cases, prepared them for.

Built

Self-paced course materials structured for autonomous progression. Information well-organized and on demand so the teacher’s role could compress into coaching, problem-solving, and personal feedback. Predictable cycles for assessment and reflection.

Outcome

Completion rates 15–25% above typical distance-education benchmarks. More importantly, the first articulation of the principle that organized everything since: when information is well-organized and on demand, human capacity can focus on coaching and problem-solving rather than content delivery.

What this proves

This is where the principle started. Everything I’ve built since works some version of that hinge.

Continue Return to the thesis  →
Working together

Three paths, depending on what you need.

Currently leading at ZKXP Innovation and running Headwater. Most project work flows through either ZKXP or Headwater, depending on whether the need is organizational AI adoption or audience/community intelligence. Here is where to go for what.

For general conversations, role inquiries, or ambiguous projects, write me directly at samcfath@gmail.com.

i.

In-house roles

For roles where AI adoption, knowledge systems, and adult-learning practice need to live in the same person.

LinkedIn →
ii.

Engagement inquiries

Project work routes through ZKXP Innovation, where the engineering, data, and AI capacity I draw on actually lives.

zkxp.xyz →
iii.

Intelligence engagements

Audience and community analysis routes through Headwater. A Creator Program for educators and online creators; Brands & Studios for enterprise.

headwater.cc →
Vancouver, BC