Software Architect
I design and build cloud platforms on AWS that scale on demand, absorb attacks and recover on their own, with hands-on work at every layer, from the frontend to the infrastructure beneath it.
Live simulation. Watch it take a traffic surge and a DDoS attack, or point at any part to see what it does.
Platforms for clinicians and patients, where protected health information has to stay safe at every step. Both were built to HIPAA requirements, with encrypted storage, role-based access and audit logging.
Prescribers needed a direct sales and fulfilment channel, and patient health information had to stay protected at every step.
The backend ran on Supabase Edge Functions, which split the platform across providers and put a ceiling on scaling ahead of user growth.
Platforms I have designed and built for very different industries, one industry at a time.
Technical Lead, Educative
A learning platform used by more than two million developers worldwide. I lead the engineering team and own architecture decisions, sprint planning and release readiness.
Learners had to be able to run code with no local setup, and many sessions had to run side by side without one affecting another.
Each learner gets a Dockerized environment that runs in the browser with zero local setup. Container isolation keeps concurrent sessions independent and reproducible across machines.
Long-running operations go through message-driven pipelines, so they never block a request and the platform stays responsive at peak traffic.
Relational and NoSQL schemas, each used where it fits, sit behind purpose-built authentication flows.
Structured logging, monitoring hooks and health checks became the team standard, cutting the time to detect and diagnose production issues.
Senior Software Engineer, Codefinity Solutions
A SaaS product that extracts rules from an organisation's policies and scores them against regulatory doctrine to surface compliance gaps.
Each customer's data had to be fully isolated, the machine-learning scoring had to be dependable, and the platform itself had to meet SOC 2, HIPAA and GDPR.
Every customer has its own PostgreSQL database. Connections are routed at runtime, and new tenants are provisioned automatically.
Uploaded policies are queued as inference jobs and processed by a model hosted on Amazon SageMaker, which extracts the rules in each policy and scores them against regulatory doctrine. Jobs are orchestrated idempotently, with retry.
Usage is metered, and costs are derived nightly against a versioned rate card.
Passwordless multi-factor sign-in with role and permission based access. The operator console sits behind step-up re-authentication and records a full audit trail.
Senior Software Engineer, Codefinity Solutions
A doctor-to-patient medication marketplace that handles thousands of patient requests every day.
Prescribers needed a direct sales and fulfilment channel, and patient health information had to stay protected at every step.
A direct sales and fulfilment channel on FastAPI, React and PostgreSQL, handling thousands of patient requests a day.
Stripe one-time and recurring billing turns single purchases into subscription revenue.
Scheduling is automated with event-driven SendGrid notifications, and patient and order data stay synchronised with third-party platforms.
Encrypted PHI storage, role-based access and audit logging are built into the platform rather than added on.
Senior Software Engineer, Codefinity Solutions
A platform for eye care clinicians that records and tracks patient consultations and surfaces clinical suggestions from them.
The backend ran on Supabase Edge Functions, which split the platform across providers and put a ceiling on scaling ahead of user growth.
The Supabase Edge Functions were migrated to AWS Lambda, consolidating the platform onto a single cloud and removing the scaling ceiling ahead of user growth.
Consultations are transcribed as they happen, which reduced note-taking time for clinicians.
HIPAA-aligned monitoring surfaces anomalies as they occur.
A Vanta integration keeps the platform audit-ready at all times.
Architecture and engineering, built from scratch
WebPanda is a multi-tenant SaaS product that helps businesses improve their local SEO. We provide local SEO services on it directly, and agencies buy it to sell those services to their own customers.
Local SEO work depends on several outside data platforms, heavy recurring jobs and regular reporting, for many agencies and their clients at once, with each one's data kept apart.
The architecture was designed for many tenants from the start. A request goes from the client to the frontend, on to the backend, and into a multi-tenant database, carrying its tenant with it the whole way. An agency is a tenant with customers of its own, which is what lets it sell the service under its own name.
Audits, rank checks and bulk updates take far longer than a request should. The backend puts them on a queue and returns straight away, and workers pick them up in the background.
The workers integrate with DataForSEO, BrightLocal, Local Dominator and other tools, and bring their data into one place per client. Their rate limits and slow responses are absorbed in the pipeline, never by a user waiting on a page.
A scheduler starts the recurring jobs and feeds them into the same pipeline, which generates a weekly and a monthly report for every client.
An agentic AI answers an operator's questions about where each client stands. It works through the backend, so it sees that client's current data and nothing outside the operator's tenant.
Requests and background jobs are logged with their tenant and job, so a failed provider call or a missing report can be traced to its cause.
Architecture and engineering
Waltz AI helps marketing agencies create custom email campaigns and manage their leads in its own built-in CRM. It also builds web pages and brand kits for them, and edits their images with AI.
Agencies run campaigns, leads, pages, brand assets and image work in separate tools. Putting them in one product means paying for AI and email on every action, so usage has to be metered and billed fairly.
Leads live in Waltz AI's own CRM, next to the campaigns sent to them. An agency builds a custom email campaign, picks its audience from the CRM and sends. The send itself is queued and delivered by background workers, and the results flow back onto each lead.
Each agency puts together its own brand kit: logos, colours, fonts and tone. Web pages are created from it and published by the platform, so what goes out matches the brand without being restyled each time.
Agencies edit their images with AI inside the product. An edit is queued, a worker sends it to the AI model, and the result returns to the agency's library when it is ready.
The platform is entirely credit-based. Every send, page and AI edit has a price in credits. Credits are reserved in a ledger before the work starts and settled when it finishes, so a failed job costs nothing and a balance can never go below zero.
Stripe handles subscriptions and plans, each of which grants credits. Agencies can also buy credits when they need more, or schedule purchases so they never run out mid-campaign. Each payment lands in the ledger as credits.
Experience shows what I have built. These show that the way I build it meets the bar AWS sets, from the fundamentals through development, operations and machine learning to architecture at the professional level.
Certified
Foundational
Cloud concepts, security, pricing and the core AWS services.
Certified
Associate
Developing, deploying and debugging applications on AWS.
Certified
Associate
Building, deploying and operating machine learning workloads.
Certified
Associate
Operating, monitoring and automating workloads on AWS.
Certified
Associate
Designing resilient, secure and cost-aware systems on AWS.
Certified
Professional
Complex, multi-account architectures and migrations at scale.
Whether you are starting from a blank page or untangling something that has grown, I can help you choose the right shape for it and build it to last.
Or a mix of them. Each is drawn with its security and monitoring in place, because neither is optional. Scroll to go through them.
One deployable application over one database, behind a firewall and a cache, with a read replica and its logs and metrics in one place. The fastest way to ship, as long as its modules are kept apart so it can be split later.
FitsA new product, a small team, a domain still taking shape.
A second pair of eyes on a system you already run: where it will break first, what it costs, and what to change in which order.
From a blank page to a design a team can build: boundaries, data, request paths, failure modes and the reasons behind each choice.
Moving a platform between providers or out of a monolith in steps, with the product staying up throughout.
Systems built to HIPAA, SOC 2 and GDPR requirements: isolation, access control, audit trails and evidence that collects itself.