Donald Pinckney
Donald Pinckney

Founder of Sectr, PhD

Sectr

I’m building Sectr, the data and coordination layer for AI agents, so teams can turn their agents into products without rewriting them.

Interests
  • Infrastructure for AI agents
  • Durable execution and serverless
  • Combining AI and static analysis
  • Photography (Flickr) and spending time with my cats.
Education

  • PhD, Computer Science, 2024

    Northeastern University

  • MS, Computer Science, 2020

    University of Massachusetts Amherst

  • The seal of the University of California, DavisThis is the seal of the University of California Davis (UC Davis) ; it contains the following text: THE UNIVERSITY OF CALIFORNIA DAVIS LET THERE BE LIGHT ; see http://commons.wikimedia.org/wiki/File:The_University_of_California_Davis.svg for more information.

    BS, Computer Science; BS, Mathematics, 2018

    University of California Davis

Download Résumé

Selected Experience

  1. Founder & CEO

      Sectr
    • Building Sectr, the data and coordination layer for AI agents. Sectr stores, streams and resumes agent conversations, so teams can turn their agents into products with no rewrite and no framework lock-in.
    Try Sectr (free beta)

  2. Senior Software Engineer (AI Foundations)

      Temporal
    • Helped build up the newly created AI Foundations team, rethinking infrastructure around the core primitives agents need to run reliably in production, from durable execution to the SDK abstractions developers build against.
    • Designed AI-native SDK primitives in Python, and a plugin system for integrating third-party agent frameworks into the Temporal SDKs.
    • Led prototyping of the integration between Temporal and AWS Bedrock AgentCore (AWS blog), and contributed to the design of the Google ADK integration.

  3. Founding Engineer

      Gitar
    • Built Jimy, the first AI coding agent which leverages deep static analysis and source code-level optimization algorithms to scale agentic AI to massive enterprise codebases while minimizing hallucinations. Gitar was later acquired by Sonar.
    • Led the AI research team on designing and implementing techniques for bridging core compiler algorithms with generative AI in Jimy.
    • Architected and implemented static analysis technologies for automated code refactoring solutions for enterprise clients, including multiple Fortune 500 companies.

  4. PhD Candidate

      Northeastern University

    • Built a new generation of intelligent package managers for JavaScript and Python based on combining deterministic constraint solving algorithms with generative AI to automatically fix common developer issues, such as solving runtime errors, patching security vulnerabilities and reducing code size.

    • Supervised and guided an undergraduate student (Federico Cassano) in building a distributed system using relational databases and container orchestration to archive every NPM package (over 36 million, 20+ TB) with low-latency (<1 min) within a large (50,000 CPU core) high-performance computing (HPC) cluster.

    • Developed a novel eval methodology (MultiPL-E) to standardize the evaluation of large language model (LLM) code generation across 19 programming languages, which is used extensively by researchers at Hugging Face, ServiceNow, IBM Research and SAP.

  5. ... and more!

Selected Projects

Building is my form of expression. Here are a selection of projects that I have worked on over the years.

Jimy
Repyro

Repyro

Combines generative AI with SAT-solving to find fixes to runtime errors caused by incompatible dependencies. Dependency management is extremely tricky to get right, as programmers need to make sure their dependency version constraints are neither too broad nor too narrow. Compounding this, if there is just one package in your whole dependency tree that has a defective constraint, then you may end up with a solution which causes bugs at runtime. To address this problem, we’ve built automated dependency repair tooling, starting with the Python/Pip ecosystem. Building on our previous work (MaxNPM), we integrate constraint solving with large language models’ suggestions for patching constraints, to iteratively narrow the search space and find a repaired dependency solution.

MaxNPM