Four systems, built end to end.
Each one started with a real constraint — permissions, cost, latency, or a process that lived in somebody's head. The case studies cover what the problem actually was, how the system is put together, and what is still missing.
01
SearchInkPrivate search over a company's own knowledge
Permission-aware retrieval with evidence-only answers
2025 — present
02
AI Task Management SystemFrom scattered messages to a ranked task queue
Rule-first extraction with an LLM fallback and 7-factor ranking
2026
03
NeLabs AI Personal AssistantAn assistant that runs continuously on real data
Gmail and Calendar connectors, retrieval, always-on deployment
2026
04
AI ThinkerEighteen AI apps, one interface language
Shared design system and dashboard shell
2026
SearchInkPrivate search over a company's own knowledge
Permission-aware retrieval with evidence-only answers
2025 — present
02
AI Task Management SystemFrom scattered messages to a ranked task queue
Rule-first extraction with an LLM fallback and 7-factor ranking
2026
03
NeLabs AI Personal AssistantAn assistant that runs continuously on real data
Gmail and Calendar connectors, retrieval, always-on deployment
2026
04
AI ThinkerEighteen AI apps, one interface language
Shared design system and dashboard shell
2026