Compare
Picking & Packing app vs CaseNotes vs Makunike.com AI Twin
The problem, what was built, the stack and the outcome, next to each other.
Clear · Up to 3 at a time
In common
- Shared stack: none in common
- Order: Picking & Packing app (2025 – 2026) → CaseNotes (Feb 2026 – present) → Makunike.com AI Twin (Oct 2026)
Swipe sideways to compare ›
Picking & Packing app ›
A Kotlin warehouse app for fulfilment teams
CaseNotes ›
South African legal research, e-discovery and learning, powered by AI
Makunike.com AI Twin ›
This website: a portfolio you can interview by phone call or chat
At a glance
- Role
- Lead, Smart Kitchen Co
- When
- 2025 – 2026
- Length
- 1 yr
- Status
- Delivered
- Type
- Mobile
- Role
- Founder / Lead Engineer
- When
- Feb 2026 – present
- Length
- 9 mo so far
- Status
- Live
- Type
- AI, Web, Mobile
- Role
- Designer & Engineer
- When
- Oct 2026
- Length
- 1 mo
- Status
- Live
- Type
- AI, Web
The problem
An Android app that drives picking and packing for warehouse and fulfilment operations.
South African legal information is scattered across portals with limited keyword search, while international e-discovery platforms are often priced beyond the local market.
Portfolios are static, but hiring conversations aren't. A recruiter wants to know whether I've shipped mobile apps, a hiring manager wants a production story in STAR form, and a sourcing agent wants structured answers it can compare. A PDF CV can't do any of that.
What was built
—
An integrated research, e-discovery and learning platform: hybrid semantic search across more than 10,000 judgments, a precedent citation graph, structured document extraction, AI-assisted discovery and stateful research workflows.
A digital version of me that holds a real interview by phone-style voice call or text, grounded only in material I curated. It changes what's on screen as it answers, logs what it can't answer so I can teach it, records and QA-scores every call, and is reachable by other AI agents over MCP and A2A.
Stack
- Kotlin
- Android
- Next.js
- FastAPI
- PostgreSQL
- pgvector
- Neo4j
- Redis/ARQ
- Azure OpenAI
- LangChain
- LangGraph
- Next.js 15
- React 19
- Tailwind CSS v4
- GPT-6.1 (Azure, Responses API)
- gpt-realtime-2.1
- WebRTC
- text-embedding-3-small
- Supabase pgvector
- TypeSafe Jev
- MCP
- A2A
- Azure Blob
- LangSmith
- Vercel
Highlights
- Led architecture and development
- Built in Kotlin for warehouse floor use
- Five-stage Graph RAG pipeline in LangGraph
- Hybrid pgvector and Neo4j retrieval, and a TAR active-learning loop
- Human-in-the-loop tool-calling chat with PostgreSQL checkpoints
- MCP server for case search and drafting
- Generative UI: the model calls typed navigate and show_panel tools, streamed to the browser as NDJSON events
- Phone-call voice on gpt-realtime-2.1 over WebRTC, with every call recorded, transcribed and QA-scored
- TypeSafe Jev in ten places: routing, tool selection, interview mode, guards, lead scoring, grounding, triage, FAQ review, agent vetting and call QA
- Self-improving RAG: weak or ungrounded answers land in a ranked CMS inbox and become FAQs
- MCP and A2A servers with request vetting and rate limits for agent-to-agent interviews
Infrastructure choices
—
- pgvector + Neo4j
- Redis / ARQ workers
- LangGraph with PostgreSQL checkpoints
- Vercel serverless (Next.js route handlers)
- Azure OpenAI
- Supabase Postgres + pgvector (HNSW)
- TypeSafe Jev
- Azure Blob Storage
Outcome
—
An integrated research and review platform grounded in more than 10,000 judgments, with a structured learning bank of 1,624 rubric-scored questions. The architecture supports resumable jobs, traceable sources and human review.
The twin never invents a fact about my career: when it doesn't know, it says so and I'm notified. Intent-matched navigation fires within about a second, before the first answer token. Every call is recorded, transcribed and scored so review starts with the calls that need it, and one knowledge base serves text, voice, MCP and A2A.