MCP server that routes AI tasks to the optimal model (Codex, GLM, Gemini, OpenRouter) using code-based complexity scoring with automatic fallback chains.
Rigorously overfitting on a learning curve
AI Engineer · Agentic AI Engineer · AI Harness Engineer
Agent orchestration, eval harnesses, tool-use pipelines — backed by 20 years of RPA/automation leadership.
Three ways to read this dossier. Pick one and the pages reorder themselves.
A twenty-year working manuscript.
Enterprise RPA and frontier AI systems, side by side. This is the working record of what actually shipped.
- 2021–present· rpa· automation· ai· leadership
- 2019–2021· rpa· automation· consulting
- 2013–2019· systems· automation· logistics· award
- 2010–2013· finance· development· systems
- 2009–2010· systems· development
- 2007–2009· web· development
- 2006–2007· development· web
- 2005–2006· systems
- 2005–2008· education· cs· research
- 2015–2015· award· recognition
Six capabilities. Connected by real projects.
Hover a node to trace which projects connect it to the rest of the stack. Each edge is a shipped project — not an aspiration.
Things I built because I had to.
Each one started as a question I couldn't stop asking — and the code is the part I still stand behind.
Open-weight LLM benchmark for sycophancy and pushback-resistance — 10 failure modes, 3-judge ensemble, paired-bootstrap CIs, per-mode robustness profiles.
AI-driven system that autonomously designs, builds, and deploys RPA workflows using multi-agent orchestration.
Agent orchestration CLI that weaves together multiple AI models into collaborative workflows.
Roo Code agentic modes — a multi-agent orchestration framework for complex coding tasks.
RAG improvement technique that adds contextual embeddings for more accurate document retrieval.
Python agent with persistent, evolving long-term memory built on Anthropic's Contextual Retrieval methodology.
FastAPI + React tool for detecting and reviewing anomalous fleet transaction data — ML-driven, ontology-backed, Supabase storage.
Still shipping.
Every commit on this portfolio, newest first. Generated from git log at build time.
- 2026-04-19c110b5cMerge pull request #4 from mlbrilliance/add-resume-pdf
- 2026-04-20fca9de5ship resume PDF so download link works
- 2026-04-19baa7107Merge pull request #3 from mlbrilliance/hook-up-cal-booking
- 2026-04-19f0abc8aMerge pull request #2 from mlbrilliance/hero-demo-more-questions
- 2026-04-19c3d747dMerge pull request #1 from mlbrilliance/favicon-refresh
- 2026-04-19400b81ewire Book-a-Call to Cal.com 30-min event
- 2026-04-19e67757badd 3 grounded demo traces (WEX, DHL award, CrewAI benefits)
- 2026-04-191f93a8aallow local static-preview origins (4321/5000/8080)
- 2026-04-19b304c9ens monogram favicon, twin photo, fix GitHub link
- 2026-04-1933a9546populate AI twin brain + split-deploy static/API, wire CORS
- 2026-04-1843fa8c3default to GLM 5.1 + add /api/health diagnostic
- 2026-04-187693c1fmigrate llm-client to OpenRouter with native model fallback
- 2026-04-18d82cdebretire HeyGen, pivot identity to AI Engineer / Agentic AI / AI Harness
- 2026-04-18da67514ambient node field + git-sourced release log
- 2026-04-18157f0e0projects as agent workspaces
- 2026-04-18616add8replace SkillGraph with React Flow CapabilityGraph
- 2026-04-18b2f4c9bhero as live agent session — scripted trace + real SSE handoff
- 2026-04-1870eca49agentic foundation — session chrome + 5 live primitives
- 2026-04-1865864ecsimplify spread layout to isLead boolean
- 2026-04-179799082hydration mismatches in Contact email + body extension attrs
- 2026-04-17a1c810aphase E — letterhead contact, ruled mode columns, enlarged skill graph
- 2026-04-17e874fd2phase D — scroll spine + custom amber cursor
- 2026-04-179d3429bphase C — editorial spread layouts for Resume and Projects
- 2026-04-172acbca5phase B — cover masthead hero + SectionOpener primitive across all sections
- 2026-04-173410d6ephase A — grain substrate + masthead chrome + editorial grid tokens
- 2026-04-17332332aGSAP stagger scroll sequences for Resume and Projects
- 2026-04-14469e885gitignore Playwright output directories
- 2026-04-14071282dend-to-end Playwright suite covering the full visitor journey
- 2026-04-14dcbce66error boundaries, SEO metadata, pluggable rate limiter
- 2026-04-145e4ab09assignment portal, contact form, email automations
- 2026-04-1480d6bf1resolve phase 1 & 2 TypeScript strict-mode errors
- 2026-04-148fcbf72AI interview twin, HeyGen avatar, project Q&A
- 2026-04-146590998core pages, design system, responsive layout
- 2026-04-1337b2d8bintelligence layer — LLM client, brain retrieval, indexing
- 2026-04-1317f89f6repository scaffold and tooling
- 2026-04-13d027b57initial commit
Ask me anything, on the record.
Pick a mode, send a message. My twin answers in my voice, grounded in my actual work.

Nick Sudh
Select a mode and start talking to my AI twin
Have something to evaluate?
Take-homes, paid trials, process audits, code reviews. Submit here; you'll get a tracking URL back.
Three ways in.
Schedule a call, send a note, or find me on the usual platforms.
- Loc
- Layton, Utah
- hello@nicksudh.com
- GH/LI
- @mlbrilliance · @nicksudh
Send a Message
Connect
Find me on the usual platforms. I read everything.