I'm Caleb Lazarus — a Solutions Architect Team Lead designing intelligent agents, workflow automation, and HubSpot AI integrations for businesses ready to operate with software, not headcount.
The shift
I help leadership teams design those systems — agents, automations, and integrations that compound, instead of dashboards that sit unread.
Capabilities
An AI-first practice grounded in 6+ years of solutions architecture. The patterns are mature; the tooling is brand new.
LLM-powered agents that handle the operational work your team keeps deferring — qualification, triage, drafting, retrieval, routing. Built on Anthropic Claude or OpenAI, instrumented from day one.
Embed AI inside HubSpot — Breeze AI, custom workflows, AI-assisted sequences, lead scoring, and content generation. Then sync it bidirectionally with Salesforce, billing, ERPs, and your data warehouse.
The connective tissue. REST & GraphQL APIs, webhook orchestration, idempotent retries, queues, and the kind of middleware that doesn't wake anyone at 2am.
HubSpot CMS, React, and headless builds with personalisation and conversion logic baked in. Engineered for Core Web Vitals, semantic SEO, and the new wave of answer-engine traffic.
End-to-end system design across CRM, marketing, sales, support, finance, and product. I lead the team that builds it — discovery through delivery, not just a diagram and a wave goodbye.
Process
No 80-page documents. No 6-month discovery. Vertical slices, shipped to production, instrumented before the next slice starts.
Stakeholder interviews, current-state map, integration audit. We name the highest-leverage workflows before any code is written.
One clear diagram. A decision log with trade-offs. Platform, AI approach, integration pattern, and measurement plan in writing.
Vertical slices to production. Each slice is real, observed, and reversible before the next one starts.
Measure. Tune prompts, raise automation coverage, retire deprecated steps. Hand over with documentation a successor team can actually use.
Clients
A selection of public case-study clients spanning financial services, energy, legal, conservation, telecoms, retail, and B2B SaaS — alongside the broader 90+ engagements I’ve led at MO.
About
I started behind a camera at AFDA, then ran digital marketing for Microsoft and Oracle accounts, and now lead a development team at MO Agency in Johannesburg. The thread has always been the same: take messy inputs, work out how the pieces fit, and ship the picture the client had in their head.
Today that picture almost always includes AI — agents that handle the operational grind, models embedded inside the CRM, retrieval systems that sit on top of internal knowledge, and automation that means people get to do the work that needs them.
MO Agency · Johannesburg
Leading a team building integrations, AI automations, and websites/web apps for enterprise clients globally.
HyperionDev
HTML, CSS, JavaScript, MySQL, MongoDB, React, Express, and HubL.
University of Cape Town
Integrated digital marketing campaign design and implementation.
Digital Kungfu · Johannesburg
Led 3 marketers running campaigns for enterprise clients including Microsoft and Oracle.
iStore · Johannesburg
Managed a team of 4 technicians; troubleshooting, servicing, and repairing Apple devices.
AFDA
Conceptual, practical, and contextual skills in film, TV, stage, and emerging media.
What I’m building toward
I want every project to ship with measurement, evals, and feedback loops baked in — so the system gets sharper after launch, not staler. Less consulting on slides. More compounding software. More AI that earns its keep on day 30, not just on day 1.
I’m looking for partners — clients, peers, founders — building seriously in this space. If that’s you, the door is open.
FAQ
A Solutions Architect designs how systems fit together end-to-end — choosing platforms, defining data flow, planning integrations, and translating business goals into technical architecture. I lead the build through delivery, not just hand off a diagram and disappear.
Discovery first. We identify the highest-leverage manual workflows and decision points before any AI work starts. Then we layer AI where it has measurable ROI — agentic automation for repetitive ops, retrieval-augmented assistants for internal knowledge, AI-assisted content and scoring inside the CRM. Always with measurement, guardrails, and human review where the stakes call for it.
Yes. I implement Breeze AI agents, content tools, and copilots, and combine them with custom AI workflows where Breeze alone doesn’t fit. I also build external AI agents that sync into HubSpot via Operations Hub and custom code actions.
Yes — bidirectional sync, custom middleware, queueing, and conflict resolution are core to the practice. I’ve shipped HubSpot ↔ Salesforce, HubSpot ↔ ERP, HubSpot ↔ data warehouse, and many bespoke integrations.
Primarily Anthropic Claude (Opus, Sonnet, Haiku) and OpenAI. The model fits the task — long-context reasoning vs. fast classification vs. tool-use orchestration. Cost, latency, and prompt caching are part of every architecture.
Both. Custom Node.js or TypeScript orchestrators when the requirement is bespoke; Anthropic’s Claude Agent SDK or OpenAI’s tools when the workflow is standard. The goal is the smallest dependency surface that meets the requirement.
I lead a team at MO Agency, so project enquiries are best routed through MO. For personal advisory, speaking, or writing enquiries, email caleb@caleblazarus.com.
Johannesburg, South Africa (UTC+2). I work remotely with clients globally and overlap comfortably with EMEA, most of APAC, and US east-coast mornings.
Contact
Send a brief on the problem, the current stack, and the outcome you’re after. I respond within one business day.
caleb@caleblazarus.com