Skip to content
Asmar.

§ 01 — Asmar Momand · Software Solution Architect · Toronto / Remote● Open to roles + engagements · 2026 Q3

AI-first & offline-first — designed as foundational architecture, not bolted on. Five platforms in development with this posture, approaching pre-launch. Working experience with LLMs, RAG, provenance, on-device inference, and the offline sync protocols that keep systems honest when the network doesn't.



I architect distributed systems, design the cloud underneath, and write the code that ships them. Solo when the team is small. With a team when it's not. The diagram above is the posture I keep returning to — distributed, event-driven, multi-tenant, pure-domain.

For founders, CTOs, and engineering leaders who need an architect who can also build: I take an idea through architecture, solution design, cloud infrastructure, implementation, and delivery — end-to-end, without handing off at the blueprint and walking away.

Twelve years across public financial management, FinTech & wealth management, insurance, and owned product. Currently a Software Solution Architect at iA Financial Group; founding Ghasi Technologies — a portfolio of five enterprise platforms — in parallel.

§ 01·5 — What I do

I take ideas
from whiteboard
to production.

Most architecture practices end at the blueprint. Mine starts there. I move with a project across the whole lifecycle — framing the idea, mapping the bounded contexts, designing the solution, standing up the cloud, and writing the code that actually ships. When the team is small, I do all of it. When it's not, I lead it.

  1. — 01

    Idea framing

    What's the real problem, and what changes if we solve it?

  2. — 02

    Architecture

    Bounded contexts, ADRs, integration patterns, risk register.

  3. — 03

    Solution design

    Service blueprints, API contracts, event schemas, data model.

  4. — 04

    Cloud infrastructure

    GCP / AWS / Azure. Networking, security, observability.

  5. — 05

    Implementation

    Production-grade code, with the team or alone.

  6. — 06

    Delivery & deployment

    CI/CD, drift gates, on-call, the boring parts that matter.

§ 01·6 — Foundational, not bolt-on

Two postures baked in. Not bolted on.

Most platforms add AI through a feature flag and call themselves AI-enabled. Most platforms add offline through a service worker and call themselves resilient. Neither is true. Both are architectural decisions that have to be made on day one — or paid for in rewrites later.

§ A

AI-first

Intelligence lives in the domain. Not in the UI. Not as a wrapper around someone else's API.

Every AI-generated artifact carries a provenance block — model, prompt-hash, tokens, cost, trace, reviewer. A single AI Gateway is the only egress for any LLM call across each platform — model selection, PHI redaction, HIPAA-mode routing, cost ceilings, and moderation are enforced at one boundary and backed by ADRs. Swapping a provider is a config change, not a code change. That's only true if it's designed in from day one.

  • Single AI Gateway in eHealth, edTech, Melmastoon, and Intake OS — the only egress for any LLM call. ABAC + API-gateway allow-list keep every other service out of the model network entirely.
  • Multi-provider LLM port — switching models is a config change, not a code change. Budget, safety, and provenance enforced at the port, not at the call site.
  • Working experience with retrieval-augmented generation, vector retrieval, tool-calling orchestration, on-device inference for offline clients, and AI evaluation pipelines.
  • AI provenance metadata is a value object rejected at the aggregate boundary when missing — the domain refuses to persist AI content without it.

§ B

Offline-first

The client is the primary actor. Network is a sometimes-optimization, not an assumption.

Clients read from local state, write to a local outbox, and reconcile with the server when connectivity allows. Conflict policies are declared per aggregate — vector clocks where order matters, last-write-wins where it doesn't. One signed sync protocol per platform; no service invents its own. Across the portfolio, this is the difference between software that works in a rural clinic, a hotel through a network outage, a field-team phone — and software that doesn't.

  • One sync protocol in edTech with per-aggregate conflict policies and an idempotent local outbox keyed by client mutation ID.
  • Vector-clock sync loop in Melmastoon's desktop back-office — designed to keep the hotel running through property-network outages without last-write-wins corruption.
  • Field-bundle BFF and offline-first desktop in eHealth — built so clinical workflows function in facilities with no connectivity and reconcile on next sync.
  • Signed, encrypted offline bundles in edTech with tamper detection — license-bound and runnable on a phone with no signal.

00+

years

across public financial management, FinTech, wealth, insurance, and owned product.

0

AI-first platforms

in the Ghasi suite — health, education, hospitality, intake, comms. Single AI Gateway per platform.

000%

offline-first

of the Ghasi platforms ship a local outbox + sync protocol. None of them assume the network is there.

0

clouds

hands-on across GCP, AWS, and Azure — architecture, infrastructure, and AI services.

§ 01·7 — Who this is for

Four audiences, one bar.

Different decisions, different time horizons, same evidence. The proof below is concrete on purpose — generic claims age badly.

For founders

One architect-builder in the first ninety days saves a team in the next twelve months.

  • Five product platforms designed and built end-to-end — health, education, hospitality, intake, comms. Every one of them AI-first and offline-first by design.
  • Idea → bounded contexts → cloud → code → CI gates → on-call. Without handing off at the blueprint.
  • 100+ bounded contexts across the portfolio; same DDD posture, different domains.
  • If your differentiation depends on AI or on working without a network, I've already built the patterns you need.

For CTOs · VPs of engineering

An architect who recommends from inside the constraint, not above it.

  • Architecture Review Board representation at iA Financial Group across FinTech / wealth / insurance.
  • Designs AI as a first-class bounded context — single gateway, provenance, budget, safety, model swap — not as a UI wrapper around an SDK.
  • Treats offline as architectural posture — local outbox, vector-clock sync, per-aggregate conflict policy — not a service-worker afterthought.
  • ADRs that name the alternatives, the tradeoffs, and the revisit conditions — written to be argued, not admired.

For engineering managers

Mentors teams. Reviews PRs. Doesn't leave the blueprint on the table and walk away.

  • Tech lead for a team of seven at iA; ran code review and PR governance.
  • Hands-on with LLM integration, RAG, on-device inference, sync protocols — your team can ask me questions in the PR, not at the architecture review.
  • Mentored juniors at Invisor; reviewed PRs across the front-end stack.
  • Mentors students through Harvard's CS50 — teaching is part of the practice.

For technical recruiters

Open to senior, staff, principal architecture roles. Toronto / remote / open to EU + MENA.

  • 12+ years across public financial management, FinTech, wealth, insurance, and owned product.
  • Deep working experience with AI-first systems (LLM integration, RAG, AI provenance, multi-provider gateways, on-device inference) and offline-first systems (sync protocols, vector clocks, signed offline bundles).
  • Trilingual: English (full professional), Pashto, Persian (both native).
  • Hands-on across all three major clouds. Certifications in progress: AZ-305 · AI-102 · SAA-C03.
§ 02 — Recent work
§ 03 — Method

Three convictions I keep returning to.

— 1 —

The domain layer is sacred.

Pure TypeScript, no framework imports, no leakage. Everything else is replaceable.

— 2 —

Multi-tenancy is not a feature flag.

It's a non-negotiable enforced at the domain, the database (RLS), and the API — or it's not multi-tenancy.

— 3 —

Architects who don't ship lose calibration.

Blueprint-only architecture decays. The discipline of running production keeps the design honest.

§ 04 — Work with me

— 01 —

Advisory

2–8 hours / week

Architecture review, ADRs, second-opinion on a roadmap. Interview support for senior hires.

Best whenyou have an instinct that something is off and need a fast calibration check.

— 02 —

Solution Design

4–8 weeks · fixed scope

Bounded-context maps, service blueprints, event schemas, ADRs, POVs and POCs, vendor evaluations.

Best whenyou know what to build and need someone to hand the team a real blueprint plus the decisions behind it.

— 03 —

Embedded Lead

1–2 quarters

Tech lead / architect role on a delivery. I own the design and supervise — or write — the build.

Best whenyou have a funded program but no senior architect on the team yet, and you want one for a fixed stretch.

Designing something hard?

I take a small number of architecture-and-build engagements per quarter, and I'm currently open to senior, staff, and principal-level architecture roles.