I help organizations redesign how work happens by combining deterministic software, AI capabilities, and human judgment into systems that are practical, durable, and built to evolve.
Three ways to improve the systems your organization already has.
01.
I analyze it, design and build an ad-hoc solution - whether that's a prompt library, a skill or subagent template, a tool, an agent, or an automation.
Delivered with full documentation and assets.
02.
I map how work actually moves through them today, improve the process based on ITIL practices, then design and build the systems - tools, agents, agentic pipelines, automations, and the logic connecting them.
Delivered with full documentation and assets.
03.
I work with you on an ongoing basis, one area at a time, to understand how work happens, define where human judgment ends and machine execution begins, and find the highest-leverage opportunities with minimal intervention and cost.
We build toward greater organizational intelligence as a continuous capability rather than a one-time project.
Open Source - Apache-2.0
Personal knowledge system that compiles, connects, and synthesizes information across sources into a living second brain. New sources continuously enrich existing knowledge instead of creating isolated notes.
Built around a compile-at-ingest architecture that uses NLP processing wherever possible and reserves LLMs for interpretation, synthesis, and ambiguity.
Accessible through wiki pages, an entity graph, built-in chat, or MCP-compatible agents.
Private / NDA
RAG-powered self-help agent for a complex retail-optimization platform with extensive configuration, rules, and decision logic. Designed to help users understand platform behavior, configure solutions, troubleshoot issues, and apply best practices without relying on support teams.
Built around a custom documentation layer and hybrid retrieval architecture. Documentation, configuration knowledge, historical cases, and operational guidance are structured, enriched, and retrieved through a purpose-built RAG pipeline that combines deterministic retrieval techniques with LLM reasoning.
Supports business and technical users through role-aware retrieval, citations, SQL assistance, and explanations grounded in platform knowledge and prior solutions.
Open Source - MIT
Agent-driven systematic literature review infrastructure that combines deterministic review workflows, LLM-assisted decision making, and human oversight to accelerate knowledge discovery and synthesis without sacrificing transparency or methodological rigor.
Deterministic components handle search, deduplication, screening workflows, tracking, and export generation, while LLMs assist with interpretation, classification, and evidence extraction. Human reviewers remain responsible for criteria, decisions, quality and final synthesis.
Produces review-ready outputs including PRISMA-compliant reporting artifacts, and downloaded open-access papers.
Personal knowledge system that compiles, connects, and synthesizes information across sources into a living second brain. New sources continuously enrich existing knowledge instead of creating isolated notes.
Built around a compile-at-ingest architecture that uses NLP processing wherever possible and reserves LLMs for interpretation, synthesis, and ambiguity.
Accessible through wiki pages, an entity graph, built-in chat, or MCP-compatible agents.
Every organization already has systems for learning, deciding, and acting. I understand how work happens today, improve what already exists, build where it creates value, and help organizations get better at it over time.
Assign each task to the cheapest layer capable of performing it reliably while preserving accountability.
01.
The backbone of reliable workflows.
Rules, calculations, retrieval, validation, and structured processes.
Low cost, consistent, auditable, repeatable.
02.
The layer that handles what rules can't.
NLP, LLMs, and machine learning for interpretation, synthesis, ambiguity, and generation.
Higher cost, flexible, context-aware, non-deterministic.
03.
The decisions that require ownership.
Taste, expectations, priorities, exceptions, and accountability.
Highest cost, cannot be fully replaced without tradeoffs.
More on my projects, building in public, and what I'm learning along the way.
https://github.com/tuirk/SLR-Engine Easy and free to run: clone the repo, open it in Claude Code (or any coding agent), and start with /slr-engine A systematic literature review (SLR) is a structured method to find, screen, and summarize published research on a…
A couple of weeks ago my bestie sent me Karpathy’s LLM wiki gist and said “I need this.” I said hang on — I’m the knowledge management girlie here. Let me build it for us. That became Kompl, the knowledge…
If you are exploring AI systems, knowledge operations, or a custom workflow build, I am happy to talk through the problem and see what would be most useful.
tuirkey 2026