Devin Gonier - Principal LLM Researcher & Applied AI Architect

summary.md

ML Research Engineer specializing in agent post-training and multi-agent reasoning systems. 5+ years leading production LLM training—GRPO, DPO, LoRA fine-tuning on distributed infrastructure—with deployment across AWS Bedrock, Modal, and vLLM. Designed multi-agent architectures that learn from both process and outcome rewards, including debate-driven systems for structured reasoning. Research published on AI pluralism and multi-agent alignment. M.S. Machine Learning, Columbia.

research_interests.md

multi_agent_systems for Reasoning and Debate
Advanced RAG.optimizeRetrieval() Techniques
knowledge_graphs and Graph-Based Learning
RL.Optimization() for LLM Alignment
Architectures for specialized_models.collaborate()

experience.json

Co-Founder & CTO

DebaterHub

Denver Jul 2021 - Present

Built voice_to_voice.AI() lessons and live debates
Agent post-training pipeline using GRPO and DPO
NSF.funded() with OpenAI and AWS grants
Multi-agent reward signals: process + outcome feedback

Principal AI Research Scientist LLMs

Arka Group

Denver Jul 2024 - Oct 2025

Led IRAD.research() for LLM applications
Built secure claude_code.alternative()
Fine-tuned LLMs using LoRA/QLoRA on A100 cluster
Designed evaluation pipelines for ISR applications

Chief AI Officer

WageUp

Denver Jan 2015 - Aug 2024

Built no_code.AI() orchestration platform for B2B
Led team of 4 devs, patents.count(4)
Designed vector_db.retrieval() and prompt optimization
Transitioned to LLM-based orchestration (2022-2024)

education.yml

M.S. in Machine_Learning

Columbia University

New York, NY Jun 2023

Relevant Coursework: NLP, Reinforcement_Learning, AI_Systems.advanced()

publications.txt

Chapter 4: Foundational Theory and Technical Grounding for AI Pluralism. Forthcoming in "AI Pluralism" (Palgrave Macmillan), Mar 2026.

A Superpersuasive Autonomous Policy Debating System. AAAI Conference, Nov 2025. DeepDebater: autonomous system for competitive policy debate.

Augmented Debate‑Centered Instruction: A Novel Research Agenda for Responsible AI Integration in Education. AAAI Conference on AI, PMLR 257:139-150, Jan 2024.

Cross Fertilizing Empathy from Brain to Machine as a Value Alignment Strategy. ArXiv, Dec 2023.

ls patents.js

AI Coach for Providing Customer Feedback to Employees — Granted #12412199, Sep 2025

Performance Evaluation Systems and Methods — #20240046194, Feb 2024

ls skills/

Technical Skills

LLM Systems & Agents

DSPy Multi-Agent Systems A2A Livekit MCP RAG ReAct Tool Use

Model Fine-Tuning & Inference

GRPO DPO LoRA/QLoRA VLLM Unsloth LLaMA-Factory PyTorch DeepSpeed

Languages & Frameworks

Python TypeScript React Vue FastAPI WebRTC WebSockets

Infrastructure & Data

AWS/Bedrock Modal Docker Weaviate Neo4j Supabase MLFlow GPU Cluster

media_links.html

TedX Talk: How to Teach Artificial Intelligence Morality and Ethics

University Of Houston AI Pluralism Expert Devin Gonier

Podcast: Singularity Syndicate: AI Pluralism, Consciousness and Emergent Properties

Using Critical Thinking to Harness the Power of AI