currently building life OS v5

I turn messy problems
into simple, useful tools.

I'm Adam — a systems-minded builder working across software, enterprise architecture, and data. I like finding the pattern, building the thing, and iterating until it's actually useful.

See life OS → Get in touch
01 / About

Where technology, data and
thoughtful design meet.

I've worked as a full-stack developer, a senior engineer on enterprise systems, and an enterprise architect shaping technical strategy for the PRESTO network.

Outside work I build tools for myself — mostly around health, fitness and personal data. I'm interested in how nutrition relates to weight, training to recovery, and circumstances to mood.

I see AI as a way to turn ideas into working systems faster, and I use it constantly.

How I work
01
Build it. A rough working version beats a perfect plan.
02
Measure it. Look for patterns in the data, not the anecdotes.
03
Be wrong, quickly. I'd rather test an assumption than protect it.
04
Simplify. Replace the messy spreadsheet with the dashboard.
02 / Experience

Where I've worked.

Enterprise Architect & Interim Manager

Apr 2025 – Present
Metrolinx
  • Drove technical strategy for the PRESTO fare-payment network.
  • Catalogued system integrations and managed cross-functional roadmaps.
  • Replaced legacy spreadsheets with Power BI dashboards.
enterprise architecturePower BIroadmapping

Senior Full-Stack Developer

May 2023 – Mar 2025
ComPeters / Urbint
  • Built scalable enterprise systems and AWS-based microservices.
  • Optimized relational database schemas.
  • Owned API integrations and data mapping.
AWSmicroservicesSQLAPIs

Full-Stack Developer (Co-op)

Jan 2022 – Aug 2022
ComPeters
  • My first professional coding job — shipped features across the stack.

Help Desk

Jan 2021 – Dec 2021
University of Guelph — CCS
  • Eight months answering tech questions by email, phone and chat; edited web pages in HTML.
03 / Featured project

life OS — structured context for an AI.

The point of life OS was never the dashboard. It's to get my whole life into a structured shape an AI can actually reason over: fitness, nutrition, sleep, health, mood, journaling and money, all in one self-hosted database on my home server. Claude reads it live through an MCP server I built, so it can analyse my data, spot patterns and suggest what to change.

Next: ground that AI in health and fitness research, so its suggestions are critical and evidence-based rather than generic.

Python ETLMCP serverClaudeHevy APIGoogle HealthPlaidself-hosted

Muscle fatigue heatmaps

Front/back body views of 7-day muscle load, unilateral lifts included.

Grapher

Plot any data point against any other, with a significance test built in.

AI daily reports

Every day gets a written report with journal context baked in.

Custom calorie model

Speed-based MET model for cycling instead of Fitbit's estimates.

Nutrition plans

Cut / Bulk / Maintain targets; history keeps the plan active at the time.

Finances & health

Accounts, categories, bloodwork and vaccine records in one place.

CHANGELOG · v1.0 May 2026 → v5.54 Nov 2026
nextA research-grounded AI analyst: trained on health and fitness research so it can critically analyse my data and make evidence-based suggestions.
v5 Sep 28Back on the home server. Plaid bank sync with rules and auto-matched transfers, rule-based workout grading (0–100), meal planning and a scaling cook mode, automations with phone alerts, Google Calendar, and custom Interests lists Claude can read and edit. Now on v5.54.
v4 Aug 29Customisable dashboards. Google Health replaces Fitbit. Biomarker charts, symptom and sleep timeline, a pantry that counts itself down, "Finish my macros" meal suggestions, and named API keys with request logs.
v3 Aug 12Renamed to life OS. Finances (1,500+ transactions from Cashew), a health form engine, muscle fatigue heatmaps, nutrition plans, multi-user accounts, and a full design system with dark mode.
v2 May 28MCP server so Claude can read live data. Daily journal with 2,141 days imported from Pixels. Speed-based MET model for cycling calories.
v1 May 19Flask + SQLite on a home-server container. Dashboard, AI daily reports, weekly progressive-overload reports.
04 / Analysis

Does salt make me heavier tomorrow?

I kept noticing my weight jump 1–2 lbs after salty days, so I tested it with my own life OS data.

Method. Daily sodium (mg) vs. the change in next-morning weight — change, not raw weight, because I was cutting from 174 to 156 lbs. Trimmed the sparse early logging, fit a linear regression, ran a one-tailed test at α = 0.05.

Result. Statistically significant but small: about +0.29 lb per 1,000 mg. Association, not causation — and 73 days is thin. Goal is 100–200+ clean days with consistent weigh-in times.

sodium (mg) → Δ weight next day (lb)
73days (n)
0.259r
0.067R²
0.013p (one-tailed)

Points are illustrative, generated to match the reported stats. Line and stats are from the real analysis.

05 / More projects

What else I'm building.

Building now

in progress

Personal enterprise architecture tool

Enterprise architecture, applied to my own life: map the systems, data and integrations I rely on the way I would for a company. A life OS integration is coming.

enterprise architecturelife OS
playable prototype

Airflow

A puzzle game about heating a house, built on a real fluid simulation. Place heaters, fans and HVAC, then watch warm air rise, drafts spill in and the sun move, while you keep every room comfortable without tripping the breaker. Next: a roguelite deckbuilder where you're a smart-home AI working through a neighbourhood.

WebGPUfluid simvanilla JSgame dev
ongoing

Home server

Proxmox hosts everything I run at home: life OS, Immich for my photo library, and Home Assistant for automations built on buttons, routines and sensors.

ProxmoxImmichHome Assistant

At work

work · PRESTO

Enterprise architecture tool rollout

Implementing an enterprise architecture tool for PRESTO, with dynamic integrations that keep it current and reporting built on top.

enterprise architectureintegrationsreporting
work

Copilot documentation assistant

An AI assistant built on Copilot that searches and retrieves answers from a large, messy documentation database, so people stop digging by hand.

CopilotAI searchknowledge management

Earlier

2023 · team of 5

Bloom Buddy — plant tracking app

A cross-platform Flutter app for plant people: a plant library, watering reminders, care reference data, and dated notes with photos to track growth over time. Designed in Figma.

FlutteriOS + AndroidFigma
2020

Star Citizen Trade Tool

The in-game route planning was lacking, so I modelled locations, resources, supply/demand and prices in a relational database and wrote a query that surfaces the most profitable surplus trades.

SQLdata modellingsimulation
06 / Contact

Let's build something useful.

Open to conversations about software, data, architecture — or your own personal dashboard.