Fast Triage
An explainable AI workflow that prioritizes a clinician's EMR inbox and evaluates referrals — built on synthetic healthcare data, with source evidence shown for every ranking and human approval at every decision point.
AI PRODUCT BUILDER × DIGITAL TRANSFORMATION OPERATOR — OTTAWA
I turn messy real-world operations into working AI-enabled systems.
SCROLL TO ORBIT
0+
SYSTEMS SHIPPED
0–0%
FASTER WORKFLOWS
0+
USER TRANSFORMATION IN ~30 DAYS
0YRS
OPERATING REAL ORGANIZATIONS
WHAT I DO — ONE OUTCOME, THREE WAYS IN
-
01
AI Product
BuildingWorking products, not prototypes — a clinical triage workflow, a board-results analytics product, a multi-tenant school ERP. I own problem selection, product judgment, and architecture; AI is the build environment.
-
02
Agentic Workflow
AutomationDirective → orchestration → execution workflows for payroll, recruitment, exam analysis, and reconciliation — 50–85% faster cycles, with a human approval gate before anything irreversible.
-
03
Digital Transformation
& Adoption2,000+ users moved to remote operations in ~30 days. ~1,000 users migrated between cloud platforms. Systems only count when people actually use them — I run the change management too.
THE METHOD — WHY MY AGENTS DON'T FALL OVER
Directive.
Orchestration.
Execution.
LLM reliability compounds badly: 90% accuracy per step decays to 59% across five steps. My 3-layer pattern splits the probabilistic from the deterministic — the agent decides, the code executes, and a human approves what matters.
-
01
Directive
Markdown SOPs written by the person who owns the process — the rules, the edge cases, the approval points. When the agent learns a new constraint, the directive is updated: the system anneals.
-
02
Orchestration
The agent reads the directives, routes the work between skills and scripts, and stops for a human before anything irreversible — bank files, bulk emails, clinical actions, record changes.
-
03
Execution
Deterministic Python and TypeScript do the actual work. No LLM in the hot path means no compounding error — the same input produces the same payroll file every single time.
FLAGSHIP WORK — BUILT & RUNNING
Things I've
shipped.
ResultRadar
Principals upload a raw CBSE board-exam gazette and get a decision dashboard — toppers, at-risk cohorts, subject-wise performance — plus an AI-written principal's report as a branded PDF.
Bright Land School ERP / HRMS
Multi-tenant SaaS platform for K-12 operations — HR, payroll, academics, fees and finance on a ~116-model Postgres schema with row-level security, plus a companion mobile app. I lived the operating context this system encodes.
Agentic Operations Systems
A family of directive → orchestration → execution workflows for payroll, recruitment, exam analysis, and student reconciliation — including an 8-stage payroll agent running three schools ~85% faster than the manual cycle.
GWS → M365 Migration
Staged migration of ~1,000 users from Google Workspace to Microsoft 365 — Azure AD sync, identity mapping, rollback planning, and a cutover nobody had to think about twice.
School Books PWA
A multi-school, multi-location bookshop sales tracker that replaced paper registers — phase one of a broader ERP rollout, installable on any device.
THE ARC
From the
ops floor
to the
agent stack.
I build AI-enabled products for messy, high-stakes operational work.
My background is unusual: I started in computer science, spent a decade operating and transforming a multi-campus education organization, and now build software and agentic workflows across education, healthcare, finance, and internal operations. My edge isn't prompting or coding — it's knowing how work actually breaks: incomplete data, policy constraints, exceptions, adoption resistance, and the need for a human approval gate. AI is my build environment; the problem selection, workflow design, architecture decisions, validation, and adoption are mine.
- 4,000+ APPLICATIONS SCREENED
- 400+ INTERVIEWS CONDUCTED
- ~100% ENROLLMENT GROWTH OVER 3 YEARS
- 20+ WORKSHOPS FACILITATED
-
2026 — NOW
AI Product Builder & Agentic Systems Designer — Independent
Shipped working AI-enabled products across healthcare, education, analytics, payroll, and operational automation — Fast Triage, ResultRadar, a multi-tenant ERP — applying the 3-layer pattern across 10+ systems with deterministic execution and human approval gates.
-
2024 — NOW
Digital Transformation Consultant — Outproduct
Discovery, implementation roadmaps, configuration, enablement, and handoff for education and SMB clients: ERP rollouts on a platform deployed across 700+ schools, a ~1,000-user cloud migration, and AI-assisted knowledge systems that cut new-staff onboarding from a week to under a day.
-
2017 — 2025
Operations & Systems → Director — Bright Land Group, India
Rose from coordinator to Director of Operations & Strategic Initiatives for a multi-campus group serving 1,500+ students and 100+ staff. Built the HR function from zero, automated payroll ~85% faster, and led the COVID remote transition for 2,000+ users in ~30 days — the operating context behind everything I now build.
-
2013 — NOW
Google Workspace Consultant — Freelance
Architecture, migrations, security configuration, automation, and training for schools and SMBs — a decade of client implementation work and trust.
-
2014 — 2015
Early Ottawa years — Espial & Grab The Map
Software developer co-op at Espial extending JavaScript middleware; co-founded an indoor-mapping startup at Carleton's Hatch incubator, leading a five-person team through product development and deployment.
MASTER OF DIGITAL TRANSFORMATION — CARLETON, EXP. 2026
B. COMPUTER SCIENCE — CARLETON, 2017
BEST APP AWARD — UOFT HACKS
DESIGN AWARD — HACKING HEALTH OTTAWA
PRESIDENT — CARLETON ROBOTICS CLUB
MENTOR — STARTUP PUNJAB
JUDGE — NASA-SUPPORTED SPACE SETTLEMENT COMPETITIONS
THE PART WHERE WE TALK
Have a manual workflow worth turning into a product? Let's talk.