Capabilities, with the evidence attached.
Every tool listed is one I’ve used in a public repository or worked through in a case study, not one I’ve only read about. Codes like CS-01 link to that work.
- CS-01 Bank policy assistant (RAG prototype)
- CS-02 Support automation (design exercise)
- CS-03 Snitch (social listening product)
- MEM Memora (memory API for AI apps)
- MCP Kubernetes MCP server
Overview: capability × project
| Capability | CS-01 | CS-02 | CS-03 | MEM | MCP |
|---|---|---|---|---|---|
| Backend systems and APIs | |||||
| Applied AI, RAG and evaluation | |||||
| Integrations and workflow automation | |||||
| Identity, authorisation and security | |||||
| Reliability, observability and system design | |||||
| Discovery and technical communication |
- 01
Backend systems and APIs
HTTP APIs, background workers and data models, written so they are easy to reason about and safe to change.
Shown inCS-01 Bank policy assistant (RAG prototype)CS-03 Snitch (social listening product)MEM Memora (memory API for AI apps)MCP Kubernetes MCP server
Tools and techniques for Backend systems and APIs, and where each was used Tool or technique Used in Python CS-01MEMMCP Flask CS-01 FastAPI MEM TypeScript CS-03 Next.js CS-03 PostgreSQL CS-03MEM Prisma CS-03 SQLite CS-01 Alembic MEM Docker Compose MEM - 02
Applied AI, RAG and evaluation
Retrieval, ranking, structured generation and LLM classification, plus the evaluation loop that keeps them honest.
Shown inCS-01 Bank policy assistant (RAG prototype)CS-03 Snitch (social listening product)MEM Memora (memory API for AI apps)CS-02 Support automation (design exercise)
Tools and techniques for Applied AI, RAG and evaluation, and where each was used Tool or technique Used in Retrieval-augmented generation CS-01 Hybrid search: FTS5 / BM25 + vectors CS-01 Reciprocal rank fusion CS-01 pgvector MEM Ollama (local models) CS-01 OpenAI embeddings MEM Structured JSON output CS-01 LLM classification with confidence bands CS-03 pypdf extraction & page-aware chunking CS-01 Human-reviewed eval cases CS-01 - 03
Integrations and workflow automation
OAuth integrations, webhooks and MCP servers that let AI take part in existing workflows without bypassing them.
Shown inCS-01 Bank policy assistant (RAG prototype)CS-03 Snitch (social listening product)MCP Kubernetes MCP serverCS-02 Support automation (design exercise)
Tools and techniques for Integrations and workflow automation, and where each was used Tool or technique Used in Microsoft Graph / SharePoint CS-01 Model Context Protocol (MCP) CS-03MCP Slack API CS-03 Linear API CS-03 Signed webhooks CS-03 Stripe CS-03 Resend (email) CS-03 - 04
Identity, authorisation and security
Permission models, scoped credentials and checks that fail closed, so AI never has more access than the person it acts for.
Shown inCS-01 Bank policy assistant (RAG prototype)CS-03 Snitch (social listening product)MEM Memora (memory API for AI apps)CS-02 Support automation (design exercise)
Tools and techniques for Identity, authorisation and security, and where each was used Tool or technique Used in Role-based access control CS-01 Delegated OAuth tokens CS-01CS-03 Scoped API keys CS-03 API-key authentication MEM CSRF protection & CSP CS-01 Fail-closed permission checks CS-01 Encrypted tokens at rest CS-03 - 05
Reliability, observability and system design
Knowing what the system did and why, and deciding deliberately how it should fail.
Shown inCS-01 Bank policy assistant (RAG prototype)CS-03 Snitch (social listening product)MEM Memora (memory API for AI apps)CS-02 Support automation (design exercise)
Tools and techniques for Reliability, observability and system design, and where each was used Tool or technique Used in Transactional outbox CS-03 Idempotency keys CS-03 Queue workers (BullMQ + Redis) CS-03 Atomic re-indexing CS-01 Audit logging CS-01CS-03 Rate limiting per principal CS-03 Retries with exponential backoff CS-03 Automated test suites CS-01CS-03MEM - 06
Discovery and technical communication
Structured discovery, written decision records, and plain-language explanations of what a system does and can’t do yet.
Shown inCS-01 Bank policy assistant (RAG prototype)CS-02 Support automation (design exercise)CS-03 Snitch (social listening product)
Tools and techniques for Discovery and technical communication, and where each was used Tool or technique Used in Discovery question sets CS-01CS-02 Decision records CS-01CS-02 Pilot & evaluation plans CS-01CS-02 Published system-design writing CS-01