I'm a senior operations professional who builds things. Not decks about building things — actual pipelines, deployed AI systems, and working tools that sit in production and do their job quietly while people get on with theirs.

My career has crossed B2B SaaS, global commodity trading, capital markets, and international development. Not because I couldn't pick a lane, but because the most interesting problems tend to ignore industry boundaries, and I've always followed the problems.

Finance degree from the University of Alberta. Fluent in English, French, and Fulani/Pular. Based in Gatineau, QC.

Claude API TypeScript Next.js Vercel

Post-call intelligence for GTM teams. Paste a sales call transcript, get back a structured deal summary, objection map, next steps, risk flags, and a ready-to-send follow-up email. Built on the Claude API with a Vercel serverless proxy.

The problem it solves: most reps spend 20–30 minutes after a call writing notes that live in a CRM field nobody reads. This cuts that to under 2 minutes and makes the output actually useful.

→ Try it live

Intercom Fin AI Experiment Design ROI Modelling Stakeholder Alignment

The one I point to when someone asks what it actually looks like when AI works in production — not as a pilot, not as a concept, but as infrastructure running quietly at scale.

Built the ROI model. Wrote the business case. Secured executive go/no-go. Designed the experiment framework, ran the controlled pilot, drove full production rollout.

Python Pandas SQL Logic Matplotlib

A three-stage pipeline that merges CRM, Support, and Product usage data to surface silent churn risk before it becomes a cancellation conversation. Built to catch the accounts that look fine on paper and aren't.

The output is a prioritized list of at-risk accounts with the specific signal that triggered the flag — not just a score, but a reason.

Python Pydantic NLP Jira API

Pipeline that reads a raw meeting transcript, extracts every action item, classifies each by type and priority, and cross-references against a revenue database to auto-escalate tickets involving at-risk enterprise clients. Ships Jira-ready JSON in under a second.

The part worth noting: a bug blocking a $120K at-risk account is not a "Medium" ticket by any honest measure. The Revenue Intelligence Gate is what makes the pipeline operationally honest.

Python NLP Classification

Rule-based AI classification of 1,000 support tickets across 7 categories at 87.2% accuracy. The case study documents the logic-ordering optimization that lifted accuracy by 11 percentage points — which turned out to be a sequencing problem, not a model problem.

ElevenLabs TTS Web Speech API Next.js TypeScript

Bilingual EN/FR voice math tutor for Grade 1–3. The browser asks the question out loud. The child answers out loud. MathVoice responds — with encouragement or gentle correction, voiced. No install, no backend, no account. Just Chrome and a kid who needs to practice their sevens.

Live demo runs on a free-tier ElevenLabs key. If the voice goes silent, credits are out — it falls back to browser TTS. Chrome only (Web Speech API).

Strategic Framework Enterprise AI Ops Playbook

A deployment playbook written from the experience of actually shipping AI in a production B2B SaaS environment. Covers risk classification, behavioral dynamics, triple-gate safety protocols, and what governance actually unlocks commercially. Less interesting than the live projects above, but more useful if you're about to deploy something and haven't thought carefully about what happens when it misbehaves.

I have a finance degree and I never fully left the field. These run on live API feeds because static data is boring.

Yield Curve Monitor — Live sovereign yield curves for US, Canada, and the Euro Area. Tracks 2s10s and 3m10y spreads, inversion signals, and recession bands across a 25-year animated timeline.

Power Market Monitor — Real-time generation mix, demand, and grid stress across ERCOT, PJM, CAISO, and AESO. The conditions that precede price spikes in physical power markets, made visible.

Both are single-file HTML tools on GitHub Pages. No frameworks, no backends, no build steps. Sometimes the best architecture is no architecture.

Senior operations, AI enablement, or product ops roles where the mandate is to build — not just advise. Most useful to organizations that are past "should we use AI?" and into "how do we do this without breaking what works?"

Drawn to high-growth SaaS, fintech, and AI-first companies where the problems are genuinely hard. Open to early and mid-stage startups with ambitious mandates.

Gatineau, QC. Remote across Canada. Open to relocation.