Independent / AI systems
Building Kikan, agent workflows, memory and orchestration experiments, scheduled automation, and local or hybrid model environments.
01 / Systems Architecture
Clinical AI across real boundaries.
Enterprise clinical-AI delivery case study.
A high-level model of delivery for large enterprise healthcare customers, from hospital triggers and de-identification to routed AI products, monitoring, and radiologist access.
Aival / Clinical-AI delivery model
Conceptual reconstruction
DICOM / CXR image objects
Patient metadata
AI output / results
Web service + access
Conceptual and confidentiality-safe: this delivery model omits client data, vendor-specific implementation, credentials, and security configuration.
02 / Product portfolio
Products are systems you can use.
One compact stage for two product modes. Scatter is a working conceptual interaction; Kikan is an early-stage working harness with interface evidence.
Read the shape of a model.
Import logits and metadata, reduce them to three dimensions, inspect classes and outliers, then select a spatial cohort for deeper review.
- Designed and implemented end-to-end by Justin.
- PCA processing + interactive OpenGL product workflow.
- Selection, gallery filtering, inspection, and CSV export.
- This visual is portfolio-specific, not historical product UI.
Kikan
An opinionated workflow that turns a one-line feature idea into researched, designed, planned, implemented, and tested change.
- Backlog, planning, development, and merge review as first-class states.
- Planner and developer context preserved in issue timelines and agent logs.
- Evidence below is from active dogfooding, not a production-autonomy claim.
One operating picture for agent work.
Action Centre, global feed, recent movements, and Kanban make the execution state legible at a glance.
Keep the reasoning trail attached to the issue.
Timeline and Agent Logs expose planner and developer context, tool calls, blocked states, and the next useful action.
Product evolution is part of the evidence.
An earlier AITT-branded dashboard iteration shows the same operating model becoming more focused and coherent.
03 / Experience
Experience.
Aival
Led a five-person engineering function and owned clinical-AI product and platform delivery from architecture through production support.
Emotech
Built Java and Spring Boot REST/OAuth services and NLP or video-recognition features for multi-region product trials exceeding 20,000 users.
Netcraft
Improved memory use in an autonomous phishing classifier by 13x and 7x across key components, while strengthening handling of adversarial browser content.
Imperial College London
MEng Computing with Artificial Intelligence, First Class Honours.
04 / Skills
Capabilities connected to evidence.
Select a category to see the systems and work that substantiate it. This is a relationship map, not a technology inventory.
Languages and application
Backend services, browser products, mobile applications, and performance-sensitive visual systems.
BackendJava, Spring Boot, Python, Node.js
Web applicationJavaScript, TypeScript, React, REST APIs, OAuth, web security
Mobile + nativeReact Native, Flutter, C++
EvidenceSpring Boot services at Emotech; React and browser automation at Netcraft; full-stack clinical products at Aival; C++ numerical work in Metalynx Scatter.
Familiar / secondaryGo, Kotlin, Rust
AI and data-intensive systems
From numerical model-space inspection to clinical-AI validation and adversarial classification inputs.
AI systemsPyTorch, active learning, explainability, medical AI validation and monitoring
Imaging + representationDICOM, PCA, 3D/OpenGL visualization
Data workflowsHigh-volume data ingestion, metadata-rich packaging, analytics workflows
Storage + querySQL, PostgreSQL, SQLite, MongoDB
EvidenceClinical-AI products at Aival; model-space analysis in Metalynx Scatter; AI research at Imperial; adversarial classification at Netcraft.
Platform and deployment environments
Systems designed around operational boundaries rather than a single idealized runtime.
Runtime environmentsGCP, Vercel, Linux, physical on-premises servers, hospital-provisioned VMs
Containers + orchestrationKubernetes, Docker, Docker Swarm
Infrastructure servicesRedis, RabbitMQ, Cloudflare Access / NAT
Delivery systemsGit, GitLab Runner, Jenkins, CI/CD
Familiar / secondaryLogstash, Kibana
Product engineering and delivery
Technical decisions remain connected to workflows, release evidence, and the people using the product.
Technical ownershipArchitecture, API design, full-stack delivery, client integration
Product deliveryProduct roadmaps, testing and release discipline, production support, remote updates
Operating contextsB2B healthcare products, regulated client deployments, B2C product launches
EvidenceArchitecture through production support at Aival; end-to-end Metalynx Scatter; multi-region launches at Emotech; research-to-test workflow in Kikan.
Leadership and high-trust operations
Team and delivery leadership grounded in architecture, release discipline, and operating context.
Team leadershipTechnical leadership, five-person engineering function, hiring, onboarding, code review
Planning + translationDelivery planning, product roadmaps, stakeholder and regulatory translation
OperationsServer hardening, backups, remote updates, production support
Regulated deliveryISO 27001, HIPAA, DTAC, GDPR
EvidenceLed Aival's engineering function while retaining hands-on ownership across architecture, delivery, client integration, and regulated operations.