Tanner McLeod
Software Engineer · AI/ML
Durham, NC
CS and Math double major at Duke building AI-powered systems — agentic tooling, retrieval pipelines, and the infrastructure underneath them. Currently a software engineering intern at Google, and researching temporal memory in adversarial reinforcement learning on the side.
B.S. Computer Science & B.S. Mathematics
Artificial Intelligence / Machine Learning concentration
Durham, NC
3.8
GPA
May 2027
Graduating
AI / ML
Focus
Double major spanning the modeling side and the systems side of machine learning — the theory that makes a model work and the engineering that makes it run.
Selected Coursework
Machine Learning
- Applied Machine Learning
- Mathematics of Machine Learning
Mathematics
- Applied Stochastic Processes
May 2026 — PresentSoftware Engineering Intern
Google · Internship
Kirkland, WA
Built an internal analytics agent that lets PMs, designers, and engineers pull feature performance insights from complex product data in natural language — adopted by a ~40-person Android product team.
- —Engineered a codebase-crawling agent orchestrating 8–12 distributed subagents, cataloging 2,000+ UI elements, 250+ telemetry counters, and 12+ analytics tables into a structured knowledge base
- —Designed a hierarchical catalog schema that expands only relevant branches, enabling efficient retrieval without overflowing the context window
- —Shipped a capability that generates pre-filtered links into the team's analytics platform, so non-technical users get the exact data cut without pinging an engineer
- —Developed LLM-as-judge evaluations gating presubmit to catch behavioral regressions
AgentsLLM EvalsPythonAnalytics
June 2025 — August 2025Software Engineer Intern
Capital One · Technology Early Internship Program
McLean, VA
Built a full-stack internal labor portal replacing an error-prone Excel/Snowflake workflow used by VPs to manage 60+ teams across 10 towers.
- —Architected serverless AWS infrastructure (Fargate, Aurora, S3) including database provisioning, networking, and dev/QA environment configuration
- —Built a multi-stage Jenkins CI/CD pipeline with Bogiefile IaC, gating deployments on SonarQube static analysis and 80% Pytest coverage
- —Implemented automatic audit logging in Aurora PostgreSQL capturing every change, author, and timestamp
ReactFastAPIAWSPostgreSQLJenkins
May 2024 — May 2025Software Development Intern
James River Equipment · Internship
Charlotte, NC
Owned data pipeline performance and enterprise access management across the customer-facing systems of a John Deere dealer network.
- —Re-engineered a multi-GB SQL ingestion pipeline, optimizing transaction boundaries and cutting execution time from several hours to 10–15 minutes
- —Led research, implementation, and deployment of Okta and Active Directory SSO across internal enterprise systems, securing access for thousands of corporate users
SQLMSSQLOktaActive Directory
Operation Blue Shell
Nature CNN agent evaluation — Mario Circuit, update 7100
Operation Blue Shell
November 2025 — PresentTemporal memory in adversarial reinforcement learning
Solo research investigating whether recurrent memory (LSTM/GRU) enables long-horizon adversarial strategies in a multi-agent SNES Mario Kart environment. Targeting a NeurIPS 2026 workshop submission.
- —Implemented a full Prioritized Fictitious Self-Play pipeline with weighted opponent sampling to compare baseline architectures against LSTM/GRU recurrency phases
- —Established IMPALA CNN as the optimal visual encoder via a 5M-step, 6-seed ablation — 67% convergence vs. 17% for Nature CNN, a 4× improvement
- —Containerized the training stack with Apptainer and deployed to Duke's SLURM cluster, scripting job submission across 8 parallel emulator instances