A working portfolio of AI-native operations.
These are real projects from inside an enterprise IT and security function, plus the tooling I build around it. Some are shipped and in daily use, some are mid-build, and some are on the roadmap. The point of each one is the same: turn AI from a demo into something the operation actually runs on.
Enterprise AI Transformation
Leading an organization-wide shift to AI as business change, not a tools rollout.
AI Transformation Program
In progressAppointed internal lead for an org-wide AI transformation under a leadership mandate. Runs the pilots, an AI steering committee, and executive quarterly reviews, framing the work as organizational change rather than software procurement.
Department AI Pilots
In progressDepartment-by-department pilots that start from a deep discovery interview, then run a fast automation track alongside a high-value knowledge-retrieval track. Legal is the flagship; Finance is mapped and queued behind it.
AI Literacy & Enablement
ShippedA full enablement stack: a four-week "AI-first" ramp for the IT team, company-wide Lunch & Learn training for 50+ staff, and a repeatable needs-collection interview playbook that turns each team's real work into a prioritized use-case list.
Local-First AI Governance
In progressA governance doctrine where running offline on local models is itself the compliance proof. Defines a three-tier participation model, a clear data-upload whitelist, and an agent-plus-skill architecture so capability compounds even as people come and go.
Automation & Agents
Cutting repetitive work to minutes, and building agents that do it unattended.
Contract Generation
In progressOne-click generation of third-party contracts from just a company name: the system pulls official business-registry data via API and fills a Word template for review, cutting a 30-minute, fully manual task to under five.
Document Pre-fill
PlannedAI extracts structured fields straight from contracts to pre-populate deal memos, so reviewers verify a draft instead of retyping schedules by hand.
Skills Library & Dev Workflow
In progressA growing library of reusable AI "skills" distilled from recurring work, each scored by how much autonomy it can safely take. Paired with a skills-driven engineering workflow that goes spec → issues → test-first build.
Self-Hosted Agent Gateway
LiveA personal AI-agent gateway wired to multiple model providers — local laptop inference, a LAN GPU box, and cloud models — with hot-swapping between them, plus a self-hosted search engine, delivering scheduled skills over chat. Includes an A/B harness to validate locally fine-tuned tool-calling models before they go live.
Job-Search Operations Pipeline
LiveAn agent-run back office for a job search: zero-cost portal scanning against ATS APIs, web-wide discovery, parallel AI evaluation of postings against a scored fit model, real-browser liveness verification, tailored CV generation, and a self-auditing application tracker — running as a daily autopilot.
Scheduled Autopilots
ShippedHeadless jobs that run unattended each morning: one sweeps the web for AI-adoption developments and writes a source-cited briefing, another scans the work stream and rewrites a living status board.
Knowledge & Retrieval Systems
Turning documents and tribal knowledge into something you can ask questions of.
Legal Knowledge RAG
In progressNatural-language Q&A over a 900-page operations manual and an IP-licensing corpus, with page-level citations and an anti-hallucination pipeline that grounds every answer in extracted source text. Runs locally so confidential material never leaves the network.
IT Department Knowledge Base
ShippedA Markdown-plus-YAML spec for any IT department: people, systems, assets, SOPs, vendors and security policy in one uniform schema that new hires self-serve and AI workflows plug into at zero cost. Ships with a validator and a worked sample.
LLM-Maintained Wikis
In progressSelf-building wikis where you drop raw sources into an inbox and an LLM incrementally synthesizes them into an interlinked, cross-referenced knowledge base. Applied to both an IT tech wiki and a personal one.
Self-Hosted Docs & RAG Stack
LiveA self-hosted documentation and retrieval platform — a collaborative wiki plus agentic document search over a semantic index — running on local GPU so sensitive documents stay in-house.
Mailbox Archive & Intelligence
LiveAn incremental pipeline that archives a corporate mailbox before retention policies purge it: pulls new inbox and sent mail, downloads every attachment, extracts and OCRs their text, AI-summarizes each one, and rebuilds a searchable knowledge-vault index — with the work calendar synced out as a final step.
Voice & Media AI
Turning hours of audio and video into accurate, structured, searchable notes.
Diarizer with Voiceprint Memory
LiveA web tool that transcribes multi-speaker audio and remembers voices: label someone once and they are auto-recognized forever, via stored speaker embeddings. Exports transcripts, subtitles and AI summaries.
Meeting-to-Notes Pipeline
LiveDrag in any recording and one call to the GPU backend runs transcode → transcription → speaker separation → voiceprint matching, then organizes it into structured notes written straight into the knowledge vault.
Media Transcription & Summary
ShippedA scheduled pipeline that turns long-form video into accurate, structured notes using local GPU Whisper, not error-prone auto-captions, plus LLM summarization. Runs overnight, fully unattended, with batch backfill.
Infrastructure & Platform
The self-hosted backbone that makes local-first, privacy-safe AI possible.
Local GPU Inference Backbone
LiveA self-hosted GPU box serving a fleet of local LLMs — text, vision and embeddings — to every other project over the network. The shared "brain" that makes everything else private by default.
Site & VPS Operations
ShippedA static personal site on a self-managed cloud VPS — nginx, automatic TLS, key-only SSH, fail2ban hardening — operated end to end from a single plain-language request.
Security Stack Virtualization
PlannedVirtualizing the security and infrastructure hardware stack to cut physical footprint by roughly 40%, pending a hypervisor decision. Part of the IT team's second-half roadmap.
Want the detail behind any of these?
Happy to walk through the architecture, the trade-offs, or the outcomes on any project here, in Vancouver or remote.
Get in touch →