Dennis – Python, React, Typescript

An emergency-department patient-flow platform — replacing a legacy MS Access system with a modern installable PWA: live ED and Urgent Care whiteboards, patient registration, triage/retriage, clinical worklists, and audit-grade patient records. The system runs cloud-primary with a warm-standby resilience model and a designed Phase-2 on-prem/HA architecture, and is timezone-correct for UTC-11 clinical wall clocks on UTC cloud infrastructure.

It was a two week project, Dennis was very fast on achieving an MVP level product due to the AI development pipelines he has in place. The idea was that the client wanted to update an old software from an older technology to a newer one.

• Architected the clinical data model as a three-tier patients → visits → encounters schema that maps 1:1 onto the legacy system for clean historical migration (~79K patient records analysed from the legacy database), separates display identity from master identity so registration corrections never rewrite medical history, and enforces MRN deduplication through partial unique indexes with a sentinel-MRN flow for unknown/unconscious patients;
• Designed a defence-in-depth write model in which the browser can never mutate clinical data directly: every write flows through audited SECURITY DEFINER Postgres RPC functions behind row-level security with restrictive MFA (AAL2) policies, producing a complete field-level, per-clinician audit trail that powers zero-click change-history views for hospital management;
• Built the identity model around real ward workflows: individually authenticated staff accounts with mandatory TOTP MFA alongside a shared clinical-workstation kiosk role — avatar lobby, shift clock-in/out, duty badges, idle timeout with view-state restore — so every action is attributed to the acting clinician while the device, not the person, holds the session;
• Built the 16-column realtime whiteboard the department runs on: Supabase Realtime push over RLS-filtered sockets with a poll fallback and staleness cue when the channel dies (wall displays never silently freeze), optimistic TanStack Query mutations for inline edits, per-clinician view persistence, and an admin-tunable SLA engine driving triage, length-of-stay, and investigation RAG cues;
• Designed and operated a governed AI-assisted delivery pipeline that is the project's defining workflow: the client's product lead prototypes features with AI coding agents on a read-only source branch, and I own the verified path to production — each slice is cherry-picked, hardened ("trust nothing"), end-state-diffed against the source, security-reviewed, migration-diffed against the live database schema, gate-swept, and browser-verified on production before merge, with per-ticket evidence trails across a 240-issue Jira project. This pipeline has shipped the entire platform surface across two ingestion waves, including a 93-ticket wave I classified as base-on-prod / amended / new by capability-auditing the live production catalogue rather than trusting commit history;
• Held the codebase to a clinical quality bar: a hard 100% test-coverage gate on statements, branches, functions, and lines (~1,400 tests), typecheck/lint/coverage/build CI on every push, CodSpeed performance benchmarking, and a full security audit driven to zero open findings (leaked-password protection, rate limiting, instant deactivation runbooks, hardened transactional email).

An interactive medical-education platform where clinicians train on AI-generated clinical case simulations — working up a differential diagnosis, ordering tests, and selecting interventions against a scored, ontology-backed engine.

• Architected the platform's medical Knowledge Graph as an isolated, read-only PostgreSQL schema (~273K rows across 39 tables) that serves as a single shared source of truth for two independent systems — the AI case generator and the case player — with deliberate access boundaries: direct server-to-server access for the generator and an internal API layer with two-tier caching and taxonomy-based progressive disclosure for the app, so the ontology cannot be mutated by its consumers and large reference tables never ship to the browser;
• Led the structural architecture, reorganising the codebase into an npm-workspaces monorepo that cleanly separates the public case-player app from an internal admin and case-authoring app behind shared, strongly-typed packages, and designed a canonical versioned clinical-case schema with a Zod-validated compilation pipeline and JSONB version history — giving the platform a single source of truth for types across both apps and safe publish/rollback of case content;
• Designed and operated an AI-first, agent-assisted content pipeline in Claude Code, where custom agent skills and a batch agent runner automatically resolve and validate medical entities against the Knowledge Graph and surface them into a human review queue, turning an open-ended manual curation task into a fast triage-and-approve workflow and removing the main bottleneck on shipping new cases;
• Improved Knowledge Graph search performance by ~99.9% (from ~450ms to under 1ms) by introducing Postgres trigram indexing and consolidating a multi-call search flow into a single database query;
• Re-architected the asynchronous case-generation pipeline (Python service → Next.js → Supabase) to be idempotent and self-healing, using signed webhooks and a reconciliation job, which resolved a silent two-week production outage and now guarantees every generation job reaches a final state;
• Identified and fixed a critical defect where every case silently scored 0/100, then built an ontology-aware scoring engine that awards partial credit using the Knowledge Graph's clinical class hierarchy across the diagnosis and intervention modules;
• Owned platform reliability and security, conducting a security audit that remediated all findings, implementing multi-tenant data isolation and tiered API rate limiting, and enforcing a multi-stage quality bar (lint, type-check, automated tests at 98% coverage, build) on every change.

A multi-tenant local-SEO and business-intelligence SaaS that consolidates AI content generation, geographic ("grid") rank tracking, Google Business Profile analytics, backlink and review monitoring, and lead generation into a single platform serving multiple client organisations.

• Built an asynchronous FastAPI backend with a Dramatiq + Redis distributed worker pool and an APScheduler-driven scheduler, offloading long-running jobs — content generation, data ETL, and rank scans — to background workers with task cancellation and Sentry error tracking;
• Built the platform's AI content-generation subsystem — OpenAI/GPT-powered article and blog generation queued through the Dramatiq worker pool to run at scale — with automated spell-checking and content-quality monitoring so AI output met publication standards before going live, then one-click publishing to Webflow and Shopify CMSs;
• Integrated Stripe subscription billing across organisation tiers and built the publishing connectors that push generated content straight into clients' live sites;
• Implemented geographic search-rank tracking on PostgreSQL + PostGIS to compute and visualise location-based ranking grids, and enforced multi-tenant isolation with JWT auth, role-based access control, and per-organisation rate limiting;
• Deployed the system on Render behind NGINX with S3-backed article storage and Redis caching, supporting both an internal operations dashboard and an external-facing client API.
• Used Claude Code as a core part of the development workflow to scaffold FastAPI endpoints, SQLAlchemy models, and Dramatiq worker tasks, with pre-commit hooks and GitHub Actions CI acting as automated guardrails on AI-generated code.

A community-driven platform for discovering and sharing family-friendly, unique local experiences. It connects travelers with authentic adventures, offering personalized itineraries, curated recommendations, and influencer-driven content.

• Built a full-stack community platform with Django/DRF and React, supporting onboarding, social features, and content discovery.
• Implemented scalable microservices with Django Channels, Redis async tasks, and optimized PostgreSQL (indexes, JSONB).
• Integrated Firebase Auth with custom JWTs for secure, stateless API access and automatic user provisioning.
• Designed CI/CD pipelines via GitHub Actions with matrix testing, automated Railway.app deployments, and staging/production workflows.
• Delivered containerized infrastructure using multi-stage Docker builds, Nginx optimization, and docker-compose with 5 services.
• Developed advanced state management with Redux Toolkit, atomic design components, and comprehensive frontend testing.
• Built 25+ REST APIs with filtering, pagination, social features, and itinerary planning under strict business rules.
• Achieved 90%+ test coverage with pytest/factory-boy for backend and React Testing Library for frontend.
• Designed scalable database with UUID keys, denormalized counts, CASCADE policies, and async updates for performance.

The next generation of Virtual Desktop Infrastructure (VDI) technology, leveraging cloud infrastructure to provide seamless and secure access to resources for thousands of internal users anytime, anywhere.

• Contributed to a specialized team focused on advancing cloud virtualization capabilities using OCI resources;
• Led the CI/CD deployment strategy, including VM provisioning, resource allocation, security automation, and virtual cloud network administration;
• Designed and implemented a major architectural redesign leveraging burstable vOCPU/OCPU technology, reducing resource costs by 40% year-over-year;
• Streamlined VM resource allocation and network configuration using Terraform;
• Automated user provisioning, application installation, and bulk VM configurations with Ansible;
• Worked with the Linux kernel to enable VM deployment on ARM and AMD architectures;
• Managed deployment automation, source control, and repository management;
• Tracked bugs, resolved issues, and implemented cost and efficiency improvements using Jira based on direct user feedback;
• Documented implementation strategies in Confluence, contributed to team knowledge sharing and participated in hiring and training new team members.

An online examination platform for K-12 education with four role-based portals — student, parent, teacher, and admin — covering test authoring, timed test-taking, automated grading, results analytics, class and enrollment management, and a shared calendar.

• Architected the data and access layer entirely on Supabase (PostgreSQL), writing SQL migrations, RPC functions, and row-level-security policies to enforce strict per-role data isolation across the four user types;
• Built the full React/Vite frontend with role-specific dashboards, a timed test-taking experience (countdown timer, progress tracking, keyboard navigation), and reporting dashboards using Chart.js and Recharts;
• Implemented PDF generation for student diplomas/certificates and teacher attestations, Google OAuth authentication, and end-to-end coverage with Playwright, deploying to Vercel with analytics and speed insights;
• Identified and designed for high-value AI capabilities on the exam data — automated question and practice-test generation from a syllabus, free-text answer grading, and personalised remediation driven by each student's performance history — structuring the tests, results, and analytics schema so these models could be plugged in cleanly;
• Built the platform with an AI-first development workflow, using LLM coding assistants to scaffold the Supabase layer (migrations, RPC functions, RLS policies), generate the role-specific React dashboards, and author the Playwright end-to-end suite.

The project is focused on developing a new SaaS platform designed to simplify financial reporting and compliance. The platform enables businesses to efficiently draft large and complex documents required for launching an Initial Public Offering (IPO) on the New York Stock Exchange, leveraging cutting-edge technology and deep industry expertise to streamline the process.

• Developed scalable REST APIs using Python, FastAPI, and Pydantic, leveraging asyncio for high concurrency for a live, multilayer document-editing platform;
• Designed and maintained cloud infrastructure on Microsoft Azure, utilizing Azure DevOps to build robust CI/CD pipelines with GitLab, Jenkins, Docker, and Docker Compose;
• Engineered efficient data management solutions with MongoDB and Cursor, ensuring high-performance storage and retrieval of data;
• Implemented comprehensive testing and monitoring frameworks using pytest and Datadog to ensure code quality and system reliability;
• Collaborated in an Agile team, utilizing Jira for task management and Confluence for process documentation to drive continuous improvement and effective project delivery.

The project focused on designing and implementing a national governmental cloud infrastructure to modernize public sector services. It included developing AI-powered chatbots to improve citizen interactions, reduce costs, and enhance efficiency across governmental departments. The solution optimized automation, data management, and accessibility, driving the digital transformation of public services.

• Led a team of 3 developers in creating scalable cloud-based SaaS platforms for government clients, ensuring efficient project delivery and strong communication within the team;
• Oversaw the full software development lifecycle, from gathering technical and business requirements to successful project implementation and maintenance;
• Delivered AI-powered solutions, including chatbots and intelligent document management systems, for both government and non-government clients;
• Managed and executed software projects for multinational corporations and government bodies, achieving significant improvements in customer retention (90%) and streamlining development processes;
• Applied Agile and Scrum methodologies to ensure project alignment with client goals and timely delivery of milestones;
• Conducted penetration testing to identify and address security issues in existing applications, mitigating OWASP vulnerabilities;
• Managed server architecture, worked with containerization technologies and cloud platform architecture.