I build software for messy, real-world operations.
I’m a data scientist and software builder for marine, energy, and retail teams. I turn the problem that’s costing you time or money into software and AI agents your people actually use.
I build software, data systems, AI agents, and analytical tools for problems that don’t have an off-the-shelf solution.
From an initial idea to a working product, I handle the engineering, modeling, and implementation. No unnecessary complexity. No elaborate strategy decks that never turn into software.
I like getting close to the hardware, running models locally, and building systems that don’t depend on someone else’s cloud. But the right tool is always the one that solves the problem.
Ways to work together
ProjectsDefined problems, working solutions.
PrototypesTest an idea before committing.
OngoingDevelopment, research, and technical advisory.
You don’t need a specification.
Tell me what’s taking too long, costing too much, or not working properly. I’ll help figure out whether software can solve it.
Grocery flyers go up Thursday. Most home cooks never line up the deals with what they actually plan to eat.
Approach
Weekly data pipeline over the flyers of 12 grocers → LLM-built recipes anchored on what’s on sale → SEO-indexed pages per store, per week, per recipe.
Result
Weekly store-specific menus with price-per-serving. A self-sustaining SEO funnel, monetized with Google AdSense and affiliate links.
Python
AWS
LLMs
Data pipelines
SEO
AdSense
Flyer to Fork · A week of dinners
Mon
$3.12/ serving
On sale
3 of 5items
Saves
$6.40vs regular
| marks a regular-price dinner, $5.80week for 4: $84
Sample menu · illustrative prices
2021–2023DS Intern → Consultant · NSCC AER
Battery Demand Reduction
Problem
Commercial buildings in Nova Scotia get walloped by demand charges. Batteries + solar can shave the peaks, if you size them right.
Approach
Time-series simulator on metered loads, NS tariff schedules, and PV generation curves. Sweeps battery/PV configurations for payback.
Result
Sizing recommendations used in active AER research and consulting engagements.
Python
Time Series
Simulation
Solar PV
Building F · 63-day analysisn = 89,281 · scrub the plot
Target
30.0%
Capacity
27kWh
Savings
$70/ mo
Backup
72.7%
Power
6kW
Cycles
20.5/ mo
2026–presentHiring manager · builder
The rest of the team
Seven AI agents with jobs, schedules, memories and performance reviews. They work out of a dedicated machine, built on OpenClaw, and they report to me.
Headcount
8
Humans
1
On shift since
Feb 2026
Hand me a messy workflow.
I’ll work out which parts need judgment, build the agent that handles them, keep your team in charge of what matters, and run it until it’s working.
Try it · pick a job you’d hand off
How I’d build itCustomer emails
Code ×2
Agent ×2
You ×1
CodeA new email arrives. Code pulls up the customer and their recent orders.
AgentReads it and sorts it: question, complaint, refund request or noise.
AgentDrafts a reply in your voice, using the order details.
YouApprove refunds, complaints and anything unusual.
CodeSends it, logs it and tags the thread.
The agent’s one job
Answer routine customer emails.
Never does alone
Refunds money or promises anything new.
How we’ll know it works
Reply time, and how often your team edits a draft.
Day one
You approve every reply.
→
Once it’s earned it
Routine answers go out on their own, once they’ve proven right.
Don’t see yours? Most workflows split the same way. Tell me about it.
Same rules, my own team
This isn’t a mock-up. Seven agents work this way for me every day, and have since February. Each has one clear job, its own tools and memory, and work that’s checked against what actually happened.
RRexIn chargeSupervision. Reviews every agent’s work, watches system health and writes the daily report.
CCarlosMarket forecasting. Researches and files forecasts that get scored against what actually happens.
JJerryEvent forecasting. Researches what’s coming up, writes forecasts and flags where it disagrees with the crowd.
PPascalApplied research. Runs experiments, collects the data and recommends which ideas deserve more testing.
AAtlasStrategy research. Builds and tests strategies on historical data, and puts forward what holds up on data it hasn’t seen.
MMarlowOpportunity scouting. Scans listings, checks them against real sale prices and recommends where to act, with its reasoning.
DDaleWeb development. Builds complete demo websites and suggests which businesses to approach.
Running since
Feb 2026
Agents
7
Model calls
67,000+
Tokens
2B+
What I can do for you
BuildDesign and build an agent system from scratch, around how your work actually runs.
RescueTake a pilot that stalled and get it into production, with the controls it was missing.
RunKeep it monitored, measured and improving, or hand it to your team with the runbook.
Data scientist and software builder for messy, real-world operations: marine, energy, and retail. I turn complex, real-world problems into reliable software, models, and dashboards that people actually use.
Role
Head of Digital & Data, GIT Coatings
Base
Halifax, Nova Scotia
Study
BSc Physics, CS minor, Dalhousie
Panel · 2024
Future of AI in transportationTransport Canada / NRCC · ITS Living Labs
I am a data scientist and software developer with 5+ years of experience building decision tools for marine, energy, and retail operations. I specialize in Python, machine learning, AI agents, time series, and modern web stacks that make insights measurable and repeatable.
I studied Physics and Computer Science at Dalhousie University. Professionally, I lead and ship analytics at GIT Coatings, where I went from Data Scientist to Vessel Analytics and Software Lead to Data Science Lead, and now run the company’s digital and data function as Head of Digital & Data. My work spans data infrastructure, forecasting, model validation, and productizing analytics for internal teams and clients.
Outside of tech, I play squash, manage and invest in real estate, and work in the garden, spending hundreds of dollars and days of my summer producing strawberries I could have bought from the grocery store for $4.
What I believe
Software should survive contact with reality.
Tools should reduce chaos, not add to it.
Good systems are built around the people who actually use them.
I care about useful, durable, practical software more than polished nonsense.
Talks and panels
Panel · 2025
AI-driven innovation in Atlantic CanadaCOVE · H₂O Conference
Investor presentation · 2025
Presenting vessel performance at GIT Coatings’ global HQ openingGIT Coatings
How the work gets built
User-centered designBuild tools around real workflows and decisions.
Measurable outcomesDefine success metrics, test against reality, iterate.
Performance and reliabilityOptimize for speed, robustness, and traceability.
Accessible by defaultClear visuals, keyboard-friendly, readable at a glance.
Tested and versionedAutomate checks, document assumptions, keep data lineage.
Experience
Period
Organization
Role
Notes
Nov 2025 – Present
GIT Coatings
Head of Digital & Data
Consulting engagement. Runs GIT's digital function end to end: data, software, analytics products, and the systems behind Advisory Services.
Aug 2024 – Oct 2025
GIT Coatings
Data Science Lead
Led the DS team. Strategic direction on analytics + ML across marine-coatings validation.
Oct 2023 – Aug 2024
GIT Coatings
Vessel Analytics & Software Lead
Headed the vessel analytics + software division. Tools for internal teams and clients.
Oct 2022 – Oct 2023
GIT Coatings
Data Scientist
Built the foundational data infra. Reporting pipelines now in daily use; validated coating efficacy.
May 2022 – Dec 2023
NSCC Applied Energy Research
Data Science Consultant
On-call DS for AER. Demand-reduction work on NS commercial buildings.
May 2021 – May 2022
NSCC Applied Energy Research
Data Science Intern
Battery + solar demand-reduction simulations under the Clean Internship Program.
May 2019 – Sep 2019
PEJK Property Services
Landscaping Manager
Ran ops on a 45-property book. Crews, scheduling, equipment.
2017 – 2022
Dalhousie University
BSc Physics, CS minor
Dean’s List. Howe Hall house president. Dalhousie IMPACT Award.
Contact
Let’s build something unglamorous.
20-minute intro call
Pick “Book a 20-min call” as the subject and I’ll reply with a few times. Replies go out within 48 hours, Mon–Fri, 9–5 Atlantic.
Privacy: Umami collects visit stats (pages, device, rough location, referrer) without cookies or stored IP addresses, kept up to six months. Form messages are used only to reply to you. Ask and I’ll delete yours.
