James Guillochon

San Diego, CA | +1 (619) 701-0974 | guillochon@gmail.com | LinkedIn | GitHub | 🎓 Google Scholar

Publications: 126 (27 first author w/ 2,579 citations, h-index 1‌4)

Education

Harvard — Postdoctoral Scholar, Astronomy — Cambridge, MA

Insight Data Science — Fellow, Data Science — Boston, MA

UC Santa Cruz — PhD, Astronomy — Santa Cruz, CA

UC Irvine — BS, Physics — Irvine, CA

Skills

Work Experience

Esri — Principal Data Scientist San Diego, CA (Home Office) - (5 years)

  • Leadership & AI Strategy: Co-led the AI & Analytics Tech Center, shaped Professional Services AI strategy and customer-readiness materials, mentored staff, and supported bids, workshops, and customer scoping.
  • Generative AI & LLM Tooling: Contributed to generative AI workflows by implementing tools that integrate large language models for image interrogation, free-form text object detection, and automated object replacement.
  • Deep Learning Imagery Analysis: Built deep learning pipelines for satellite and aerial imagery covering zero-shot detection, land classification, algal-bloom prediction, feature extraction, and time-series monitoring.
  • 3D Scene & LIDAR Optimization: Optimized 3D LIDAR processing pipelines and applied PointCNN models alongside advanced AI algorithms and image embeddings to identify infrastructural assets scenes.
  • Agentic AI Harnesses: Developed multiplatform agentic harnesses with MCP servers and CLIs for model serving, image AI proxying, and tool orchestration across local and cloud developer workflows.
  • Oriented Imagery Asset Extraction: Delivered oriented-imagery and embedding-based tools that extract infrastructural assets from street-level and oblique imagery to support transportation and utility workflows.
  • Semantic Imagery Search: Built embedding-based search tools that retrieve similar satellite scenes so analysts can find relevant imagery without training a new detector for each request.
  • Scanned Document Asset Extraction: Built reusable document AI tools that extract assets and mapped extents from scanned records, combining OCR, language models, and geospatial enrichment for customer delivery.
  • Automated Anomaly Detection: Designed and deployed automated solutions to actively monitor open-source map catalogs and identify suspicious changesets.
  • PS AI Enablement Tooling: Built reusable catalogs and assistants for staff skills discovery, engagement battle cards, code-sample reuse, and customer-needs drafting to accelerate proposal and delivery work.

Berkshire Grey — Principal Data Scientist Lexington, MA - (2 years)

  • Robotics Software Stack: Led optimization efforts of Python/ROS software stack used for pick and place in a multi-component robotics system, including robotic arms, high-performance conveyors, and automated shuttles.
  • Motion Planning & Parameter Prediction: Created and integrated parameter prediction to select optimal robotic motions for unseen products on demand.
  • Bayesian Sensing & Pick Verification: Developed online Bayesian analysis of data from sensors integrated into the picking system to dynamically determine how many items the robotic arm lifted.
  • Operations Analytics: Curated a dashboard of key performance indicators (KPIs) for the deployed robotic system via the ELK Stack.

Harvard-Smithsonian — NASA Einstein/ITC Fellow Cambridge, MA - (5 years)

  • Research Leadership: Technical lead of an eight-person research team of undergraduate/graduate students and postdocs resulting in dozens of completed scientific projects with 10k+ citations.
  • Science Communication: Appeared on the NOVA television program in 2018, describing data analysis by my group to measure a black hole’s mass using a recently destroyed star.
  • MOSFiT & Bayesian Modeling: Developed MOSFiT, a Bayesian engine to infer the most likely analytical models for time-series data.
  • Open Astronomy Catalogs: Led creation of the Open Astronomy Catalogs, a platform for sharing astronomical data, with a user-facing frontend and an API backend to serve data quickly to users.