Skip to content
Colin M. Skinner, PhD

About

Headshot of Colin M Skinner, PhD

Colin M. Skinner, PhD

Bioengineering, UC Berkeley

I solve complex problems with software, machine learning, and AI agents.

My background spans software engineering, machine learning, and computational biology. I earned my PhD in Bioengineering from UC Berkeley, with a Designated Emphasis in Computational Biology, where my research focused on machine learning methods for DNA methylation clocks and aging biology.

My recent work has focused on AI systems and developer infrastructure, including multi-agent orchestration, research platforms, retrieval systems, and tools for building reliable AI workflows. I'm especially interested in problems at the intersection of machine learning and software engineering.

Resume

Experience, education, and skills

Software engineer and Ph.D.-trained computational biologist building AI-powered software systems spanning backend services, cloud infrastructure, semantic and graph retrieval, multi-agent orchestration, and reproducible evaluation. Built a research platform indexing 35M+ publications and pipelines generating millions of benchmark examples.

Download PDF

Experience

January 2025 – August 2026

Software Engineer — Agents & Tooling Team

SpaceXAI

  • Architected and built Astral, an internal AI-assisted research platform for the Human Data team centered on a corpus of 35M+ scientific publications, integrating semantic search, knowledge discovery, collaborative research tools, and agent-assisted workflows.
  • Built cloud-native ingestion and enrichment pipelines using OpenAlex, Crossref, LLM-assisted metadata extraction, vector embeddings, PostgreSQL, Cloud Storage, and Neo4j to continuously process and index scientific literature.
  • Designed and implemented a deterministic genomics benchmark pipeline that generated millions of reproducible evaluation question-answer pairs through configurable DNA mutation simulation, protein translation, and HGVS annotation, replacing a proposed LLM-based workflow.
  • Developed literature retrieval and knowledge-discovery systems combining natural-language search, semantic similarity, graph analytics, citation analysis, PGVector, Neo4j, and Louvain community detection.
  • Built collaborative PDF annotation and citation-aware research workflows that synchronized highlights into structured notes and enabled AI agents to organize and expand research documents.
  • Engineered a hierarchical Director/Project Manager/Worker multi-agent orchestration platform with a custom episodic-memory layer, enabling concurrent management of up to 14 engineering projects.
  • Built and operated secure production infrastructure on Google Cloud Platform using Docker, Cloud Run, Compute Engine, Cloud SQL, Secret Manager, IAP, load balancing, FastAPI services, and GitLab CI/CD.
  • Supported a roughly 200-person internal research organization, with Astral reaching 40–50 weekly active users and becoming a default starting point for literature discovery and dataset creation.

Selected projects

  • Aria | Independent Multi-Agent Engineering Orchestrator

    Built cross-model agent collaboration across coding assistants using shared chat, agent-to-agent messaging, independent terminal interfaces, and role-based approval gates for implementation and testing. Implemented episodic memory distillation from completed task cards and two proposal-based improvement loops: learning from past execution and discovering new agent methods from external research, with human approval. Created a voice-enabled mobile progressive web app accessed securely through Tailscale for remote task launch, status updates, and workflow control.

  • AI News Intelligence Platform | colinmskinner.com

    Built a news ingestion and filtering pipeline across 16 sources, with cached headlines and summaries powering an interruptible real-time voice briefing agent that supports contextual questions, web search, and resuming its scripted brief. Developed embedding-based topic visualization using UMAP and momentum comparisons across 24-hour, 3-day, 7-day, 30-day, and 90-day windows to surface emerging, sustained, and declining trends.

Education

  1. 2025

    Ph.D., Bioengineering (Computational Biology)

    University of California, Berkeley / University of California, San Francisco

    NIH T32 Predoctoral Fellow

  2. 2017

    B.S., Biomedical Engineering

    University of California, Berkeley

    Jack Kent Cooke Scholar

Skills

Languages
Python, SQL, Bash
AI & ML
PyTorch, Hugging Face, LLM APIs, Multi-Agent Systems, AI Evaluation, Embeddings, Vector Search, Semantic Retrieval, Knowledge Graphs
Backend & Data
FastAPI, REST APIs, PostgreSQL, PGVector, Neo4j, Cloud SQL
Cloud & Infrastructure
Google Cloud Platform (Cloud Run, Compute Engine, Cloud SQL, Cloud Storage, Secret Manager, IAP, Load Balancers), Docker, Linux, GitLab CI/CD
Frontend
Next.js, React, TypeScript, HTML, CSS
Developer Tools
Git, GitHub, GitLab, VS Code, Claude Code, Codex, Grok Build, MCP Servers, Jupyter

Honors

  • xAI STEM Tutor Employee of the Month (2026)
  • NIH T32 Predoctoral Fellowship
  • Jack Kent Cooke Scholarships

Dissertation

PhD, Bioengineering, UC Berkeley, 2025

Hypothesis-Driven Training for DNA Methylation Clocks and Feature Rectification for Linear Model Coherence to Detect Inflammaging

Research

Publications

Works listed on ORCID 0000-0002-6335-4493. Preprints are labeled as such.

ORCID 0000-0002-6335-4493 (opens in a new tab)

Now

Shipping Aria, a powerful coding agent orchestrator, this site's own AI News aggregator as a small product and running stephenkingtruevalue.com as a free collector resource. Open to conversations about product, platform, and research-adjacent engineering roles.

Interests

I enjoy building agents, tools, and software that help people solve real problems. I'm particularly interested in applying advances in AI and machine learning to create practical, useful systems.

My broader interests lie at the intersection of mathematics, computer science, and scientific discovery. I'm fascinated by the mathematical foundations of machine learning, particularly objective functions, interpretability, and the limitations of existing approaches. I'm driven by a desire to push the frontiers of our understanding of computation and learning, and to develop AI systems that do more than solve problems: systems that help us formulate new scientific theories and uncover fundamental principles governing the universe.

Contact

The contact form sends a message to me. Email works too.