Dalton Payne

[email protected] | linkedin.com/in/daltonpayne | daltonpayne.ai

Summary

AI engineer building production agent systems for regulated industrial domains — nearly two years shipping AI agents across Drilling, Completions, and Safety at Devon Energy, now Member of Technical Staff at Collide building the agent layer of an industrial AI platform.

Experience

Collide — Member of Technical Staff (Remote)

January 2026 – Present

  • Shipped the platform's subagent pattern — declarative templates discovered via Python entry points with a transactional instantiation service — and built primitive subagents on it over tools authored by domain experts.
  • Owned an end-to-end domain workflow integration: ported a standalone public-regulatory-data app into the platform across five services with a coordinated schema migration and agent-tool registration. Set the pattern for subsequent workflows.
  • Extended document ingestion to industrial formats (LAS well logs, Outlook .msg, .pptx) and replaced direct cloud-storage URLs with an authenticated streaming proxy across three services, closing a token-leak risk.
  • Built and maintained curated evaluation datasets across multiple production agents using LangSmith; ran iteration cycles surfacing drift between model and prompt changes.

Devon Energy — Oklahoma City, OK

June 2022 – January 2026 (3 yrs 8 mos)

AI Engineer (June 2024 – January 2026)

  • Lead engineer for production AI agents serving Drilling, Completions, and Safety in a regulated, high-consequence environment — RAG knowledge-base plugin + tool-calling against Snowflake, with versioned prompts and curated eval datasets.
  • Presented "Leveraging AI Agents to Navigate Safety Data" at the 2025 UTA Oil & Gas Conference.
  • Designed a multi-agent NL-to-SQL workflow — 6 task-specialized agents with task-tuned temperatures, ported from n8n to Python (pydantic-ai) for reuse.
  • Built a production ETL + analytics dashboard for foreman field audits on Snowflake, Databricks, Azure OpenAI (gpt-4o), and Dash Enterprise — incremental processing with threshold gating and LLM caching.
  • Authored a shared agent template + prompt-generator adopted across D&C agents; mentored an intern and taught a recurring "Teacher Thursday" session for engineers and leadership.

Data Science Intern (August 2022 – May 2024)

  • Built data pipelines and ML models on Databricks against Snowflake and OSIsoft PI, including a predictive maintenance model for an artificial lift system, scoped with petroleum and operations SMEs.

Selected Projects

kalashnikov.ai — Founder & Sole Engineer (2026 – Present)

  • AI reference platform with an autonomous extraction pipeline producing passage-level citations; full ownership on Cloudflare D1, Workers, and R2.

Technical Skills

  • Agent engineering: LangChain ecosystem, tool-calling, RAG, subagent patterns, eval datasets and iteration loops, MCP, prompt versioning
  • Data systems: Snowflake, Databricks, Alation, OSIsoft PI
  • Languages & frameworks: Python (FastAPI, pydantic-ai, pandas, scikit-learn), SQL, TypeScript
  • Cloud / Infra: Azure (OpenAI, Functions, APIM), Cloudflare, Docker
  • Certifications: FAA Part 107 Commercial Drone Pilot (2024)

Education

B.S. Data Science, University of Central Oklahoma — May 2024 (GPA 3.68)

Awards

  • Leonard-Murray Statistics Endowed Award (2023) and John Taylor Beresford Endowed Scholarship for Computer Science (2022) — UCO's largest STEM scholarship awards.