Christine Straub

AI & Machine Learning Tech Lead | LLMs | Generative AI | Agentic AI | MLOps

San Clemente, California, United States

About

Lead Machine Learning/AI Engineer with 8+ years building production AI systems across defense, healthcare, and enterprise software. Working on Reinforcement Learning environment curation and agent training, GRPO-based agent training pipelines, difficulty-calibrated Reinforcement Learning environment design, and data curation at scale. Deep experience in LLMOps, MLOps, Evaluation Pipelines, Multi-agent orchestration, and VLM pipelines across HIPAA and DoD environments. Core Expertise: - RL Environment Curation - Agent Training & Evaluation - Data Curation at Scale - Web-Scale LLMOps - Document Intelligence - LLM Orchestration - Evaluation Pipelines - RAG ๐Ÿ”— Portfolio: https://christinemstraub.com ๐Ÿ“‚ GitHub: https://github.com/christinemstraub ๐Ÿ“ Notion: https://christinestraub.notion.site/ml

Experience

  • Lead AI Engineer, Reinforcement Learning at A.Team
    Feb 2026 - Present ยท 5 mos

    - Design and build RL environments and reward functions that produce clean, learnable signals for frontier models on ML research and engineering tasks. - Build deep expertise across the frontier of ML research, training, and inference infrastructure. - Collaborate with others to brainstorm and create new ideas and tools to improve the environment-building process

  • Senior Machine Learning Engineer at Medici Land Governance Inc.
    Jul 2025 - Present ยท 1 yr

  • Senior Machine Learning Engineer at RIOS Intelligent Machines
    Nov 2024 - Jul 2025 ยท 9 mos

    โ€ข ๐—•๐˜‚๐—ถ๐—น๐˜ ๐—ฝ๐—ฟ๐—ผ๐—ฑ๐˜‚๐—ฐ๐˜๐—ถ๐—ผ๐—ป ๐— ๐—Ÿ๐—ข๐—ฝ๐˜€ ๐—ถ๐—ป๐—ณ๐—ฟ๐—ฎ๐˜€๐˜๐—ฟ๐˜‚๐—ฐ๐˜๐˜‚๐—ฟ๐—ฒ using ๐— ๐—ฒ๐˜๐—ฎ๐—ณ๐—น๐—ผ๐˜„ workflows and ๐—ž๐˜‚๐—ฏ๐—ฒ๐—ฟ๐—ป๐—ฒ๐˜๐—ฒ๐˜€ for ๐—ฌ๐—ข๐—Ÿ๐—ข-๐—ฏ๐—ฎ๐˜€๐—ฒ๐—ฑ ๐—ฐ๐—ผ๐—บ๐—ฝ๐˜‚๐˜๐—ฒ๐—ฟ ๐˜ƒ๐—ถ๐˜€๐—ถ๐—ผ๐—ป ๐—ฝ๐—ฟ๐—ผ๐—ฐ๐—ฒ๐˜€๐˜€๐—ถ๐—ป๐—ด. ย ย โ€ข Refactored standalone scripts into structured ๐—ฝ๐—ฎ๐—ฟ๐—ฎ๐—น๐—น๐—ฒ๐—น ๐—ฝ๐—ฟ๐—ผ๐—ฐ๐—ฒ๐˜€๐˜€๐—ถ๐—ป๐—ด ๐—ฎ๐—ฟ๐—ฐ๐—ต๐—ถ๐˜๐—ฒ๐—ฐ๐˜๐˜‚๐—ฟ๐—ฒ with FrameServerHandler. Reduced processing latency by 75% with real-time performance optimization. ย ย โ€ข Developed "Thoth v0" ๐—ฑ๐—ฎ๐˜๐—ฎ ๐˜€๐—ฐ๐—ถ๐—ฒ๐—ป๐—ฐ๐—ฒ ๐—ฝ๐—น๐—ฎ๐˜๐—ณ๐—ผ๐—ฟ๐—บ with bidirectional ๐—™๐—ถ๐—ณ๐˜๐˜†๐—ข๐—ป๐—ฒ-๐—˜๐—ป๐—ฐ๐—ผ๐—ฟ๐—ฑ ๐—ถ๐—ป๐˜๐—ฒ๐—ด๐—ฟ๐—ฎ๐˜๐—ถ๐—ผ๐—ป for ๐—บ๐—ฎ๐—ฐ๐—ต๐—ถ๐—ป๐—ฒ ๐—น๐—ฒ๐—ฎ๐—ฟ๐—ป๐—ถ๐—ป๐—ด ๐˜„๐—ผ๐—ฟ๐—ธ๐—ณ๐—น๐—ผ๐˜„๐˜€. Created 5-step automated ๐—ฎ๐—ป๐—ป๐—ผ๐˜๐—ฎ๐˜๐—ถ๐—ผ๐—ป ๐—ฝ๐—ถ๐—ฝ๐—ฒ๐—น๐—ถ๐—ป๐—ฒ from data sampling to model retraining. Accelerated Machine Learning cycles by 65% using ๐—”๐—ช๐—ฆ ๐—ฆ๐Ÿฏ and ๐— ๐—ผ๐—ป๐—ด๐—ผ๐——๐—• backend. ย ย โ€ข ๐—˜๐—ป๐—ด๐—ถ๐—ป๐—ฒ๐—ฒ๐—ฟ๐—ฒ๐—ฑ ๐—ณ๐—ฎ๐—ฐ๐˜๐—ผ๐—ฟ๐˜† ๐—ฝ๐—ฎ๐˜๐˜๐—ฒ๐—ฟ๐—ป ๐—ฑ๐—ฎ๐˜๐—ฎ ๐—ฎ๐—ฟ๐—ฐ๐—ต๐—ถ๐˜๐—ฒ๐—ฐ๐˜๐˜‚๐—ฟ๐—ฒ with ๐—ฑ๐˜†๐—ป๐—ฎ๐—บ๐—ถ๐—ฐ ๐—ฏ๐—ฎ๐˜๐—ฐ๐—ต๐—ถ๐—ป๐—ด for sequential video processing. Implemented sequence_id-based organization with timestamp ordering for ๐˜๐—ถ๐—บ๐—ฒ-๐˜€๐—ฒ๐—ฟ๐—ถ๐—ฒ๐˜€ ๐—ฑ๐—ฒ๐—ฒ๐—ฝ ๐—น๐—ฒ๐—ฎ๐—ฟ๐—ป๐—ถ๐—ป๐—ด models. ย ย โ€ข ๐—ข๐—ฝ๐˜๐—ถ๐—บ๐—ถ๐˜‡๐—ฒ๐—ฑ ๐—ฑ๐—ฎ๐˜๐—ฎ ๐—ฝ๐—ถ๐—ฝ๐—ฒ๐—น๐—ถ๐—ป๐—ฒ performance from days to hours for robotics applications. ย ย โ€ข ๐—–๐—ฟ๐—ฒ๐—ฎ๐˜๐—ฒ๐—ฑ ๐—ฐ๐˜‚๐˜€๐˜๐—ผ๐—บ ๐—”๐—œ/๐— ๐—Ÿ ๐˜๐—ผ๐—ผ๐—น๐—ถ๐—ป๐—ด including "Thoth Frames Retriever" FiftyOne plugin with multi-camera support and ๐—š๐—ฃ๐—จ-๐—ฎ๐—ฐ๐—ฐ๐—ฒ๐—น๐—ฒ๐—ฟ๐—ฎ๐˜๐—ฒ๐—ฑ ๐—ฝ๐—ฟ๐—ผ๐—ฐ๐—ฒ๐˜€๐˜€๐—ถ๐—ป๐—ด. ย ย โ€ข Built 5 core computer vision operators using VOXEL51 Plugins for model deployment and segmentation training. Improved team productivity by ~50% through cross-functional collaboration. ย ย โ€ข ๐—”๐—ฟ๐—ฐ๐—ต๐—ถ๐˜๐—ฒ๐—ฐ๐˜๐—ฒ๐—ฑ ๐—ฑ๐—ถ๐˜€๐˜๐—ฟ๐—ถ๐—ฏ๐˜‚๐˜๐—ฒ๐—ฑ ๐˜€๐˜†๐˜€๐˜๐—ฒ๐—บ๐˜€ with ๐——๐—ผ๐—ฐ๐—ธ๐—ฒ๐—ฟ containerization and ๐˜ƒ๐—ถ๐—ฑ๐—ฒ๐—ผ ๐˜€๐˜๐—ถ๐˜๐—ฐ๐—ต๐—ถ๐—ป๐—ด capabilities for ๐—ฒ๐—ฑ๐—ด๐—ฒ ๐—”๐—œ ๐—ฎ๐—ฝ๐—ฝ๐—น๐—ถ๐—ฐ๐—ฎ๐˜๐—ถ๐—ผ๐—ป๐˜€. Implemented ๐—ต๐—ถ๐—ด๐—ต-๐—ฝ๐—ฒ๐—ฟ๐—ณ๐—ผ๐—ฟ๐—บ๐—ฎ๐—ป๐—ฐ๐—ฒ ๐—ฐ๐—ผ๐—บ๐—ฝ๐˜‚๐˜๐—ถ๐—ป๐—ด with configurable parameters and ๐—ฐ๐—ผ๐—บ๐—ฝ๐—ฟ๐—ฒ๐—ต๐—ฒ๐—ป๐˜€๐—ถ๐˜ƒ๐—ฒ ๐—น๐—ผ๐—ด๐—ด๐—ถ๐—ป๐—ด.

  • Senior Machine Learning Engineer at unstructured.io
    May 2023 - Apr 2025 ยท 2 yrs

  • Senior Software Engineer Technical Lead ( AI / ML ) at Sapient Logic LLC
    Jul 2023 - Feb 2024 ยท 8 mos

    ๐—Ÿ๐—ฒ๐—ฑ ๐—ฑ๐—ฒ๐˜ƒ๐—ฒ๐—น๐—ผ๐—ฝ๐—บ๐—ฒ๐—ป๐˜ ๐—ผ๐—ณ ๐˜๐˜„๐—ผ ๐—ฐ๐—ฟ๐—ถ๐˜๐—ถ๐—ฐ๐—ฎ๐—น ๐——๐—ฒ๐—ฝ๐—ฎ๐—ฟ๐˜๐—บ๐—ฒ๐—ป๐˜ ๐—ผ๐—ณ ๐——๐—ฒ๐—ณ๐—ฒ๐—ป๐˜€๐—ฒ ๐—”๐—œ ๐—ฝ๐—ฟ๐—ผ๐˜๐—ผ๐˜๐˜†๐—ฝ๐—ฒ๐˜€ (๐—ฃ๐—ข๐—Ÿ๐—”๐—ฅ & ๐—–๐—ข๐— ๐—˜๐—ง), ๐—ฑ๐—ฒ๐—น๐—ถ๐˜ƒ๐—ฒ๐—ฟ๐—ถ๐—ป๐—ด ๐—ฝ๐—ฟ๐—ผ๐—ฑ๐˜‚๐—ฐ๐˜๐—ถ๐—ผ๐—ป-๐—ฟ๐—ฒ๐—ฎ๐—ฑ๐˜† ๐—บ๐—ฎ๐—ฐ๐—ต๐—ถ๐—ป๐—ฒ ๐—น๐—ฒ๐—ฎ๐—ฟ๐—ป๐—ถ๐—ป๐—ด ๐˜€๐˜†๐˜€๐˜๐—ฒ๐—บ๐˜€ ๐˜๐—ต๐—ฎ๐˜ ๐—ฒ๐—ป๐—ต๐—ฎ๐—ป๐—ฐ๐—ฒ๐—ฑ ๐—บ๐—ถ๐—น๐—ถ๐˜๐—ฎ๐—ฟ๐˜† ๐—ถ๐—ป๐˜๐—ฒ๐—น๐—น๐—ถ๐—ด๐—ฒ๐—ป๐—ฐ๐—ฒ ๐—ผ๐—ฝ๐—ฒ๐—ฟ๐—ฎ๐˜๐—ถ๐—ผ๐—ป๐˜€ ๐—ณ๐—ผ๐—ฟ ๐Ÿญ๐Ÿฌ,๐Ÿฌ๐Ÿฌ๐Ÿฌ+ ๐—ฎ๐—ฐ๐˜๐—ถ๐˜ƒ๐—ฒ ๐˜‚๐˜€๐—ฒ๐—ฟ๐˜€ ๐—ฎ๐—ฐ๐—ฟ๐—ผ๐˜€๐˜€ ๐˜๐—ฎ๐—ฐ๐˜๐—ถ๐—ฐ๐—ฎ๐—น ๐—ฒ๐—ป๐˜ƒ๐—ถ๐—ฟ๐—ผ๐—ป๐—บ๐—ฒ๐—ป๐˜๐˜€. ๐—ฃ๐—ข๐—Ÿ๐—”๐—ฅ - ๐— ๐—ผ๐—ฏ๐—ถ๐—น๐—ฒ ๐——๐—ผ๐—ฐ๐˜‚๐—บ๐—ฒ๐—ป๐˜ ๐—œ๐—ป๐˜๐—ฒ๐—น๐—น๐—ถ๐—ด๐—ฒ๐—ป๐—ฐ๐—ฒ ๐—ฆ๐˜†๐˜€๐˜๐—ฒ๐—บ ๐—”๐—ฟ๐—ฐ๐—ต๐—ถ๐˜๐—ฒ๐—ฐ๐˜๐—ฒ๐—ฑ ๐—ฒ๐—ป๐—ฑ-๐˜๐—ผ-๐—ฒ๐—ป๐—ฑ ๐—บ๐—ผ๐—ฏ๐—ถ๐—น๐—ฒ ๐—ข๐—–๐—ฅ ๐—ฝ๐—ถ๐—ฝ๐—ฒ๐—น๐—ถ๐—ป๐—ฒ processing 3+ languages (Arabic, Chinese, French) with 95%+ accuracy for tactical document capture in ๐—ป๐—ฒ๐˜๐˜„๐—ผ๐—ฟ๐—ธ-๐—ฑ๐—ฒ๐—ป๐—ถ๐—ฒ๐—ฑ ๐—ฒ๐—ป๐˜ƒ๐—ถ๐—ฟ๐—ผ๐—ป๐—บ๐—ฒ๐—ป๐˜๐˜€. ๐—œ๐—บ๐—ฝ๐—น๐—ฒ๐—บ๐—ฒ๐—ป๐˜๐—ฒ๐—ฑ ๐—ฟ๐—ฒ๐—ฎ๐—น-๐˜๐—ถ๐—บ๐—ฒ ๐— ๐—Ÿ ๐˜๐—ฟ๐—ฎ๐—ป๐˜€๐—น๐—ฎ๐˜๐—ถ๐—ผ๐—ป ๐˜€๐˜†๐˜€๐˜๐—ฒ๐—บ using ๐—š๐—ผ๐—ผ๐—ด๐—น๐—ฒ ๐— ๐—Ÿ ๐—ž๐—ถ๐˜ and ๐—ง๐—ฒ๐˜€๐˜€๐—ฒ๐—ฟ๐—ฎ๐—ฐ๐˜๐Ÿฐ๐—”๐—ป๐—ฑ๐—ฟ๐—ผ๐—ถ๐—ฑ, reducing ๐—ณ๐—ผ๐—ฟ๐—ฒ๐—ถ๐—ด๐—ป ๐—น๐—ฎ๐—ป๐—ด๐˜‚๐—ฎ๐—ด๐—ฒ ๐—ฑ๐—ผ๐—ฐ๐˜‚๐—บ๐—ฒ๐—ป๐˜ ๐—ฝ๐—ฟ๐—ผ๐—ฐ๐—ฒ๐˜€๐˜€๐—ถ๐—ป๐—ด time from hours to seconds. ๐——๐—ฒ๐˜€๐—ถ๐—ด๐—ป๐—ฒ๐—ฑ ๐˜€๐—ฒ๐—ฐ๐˜‚๐—ฟ๐—ฒ ๐—ฐ๐—ฟ๐—ผ๐˜€๐˜€-๐—ฑ๐—ผ๐—บ๐—ฎ๐—ถ๐—ป ๐—ฑ๐—ฎ๐˜๐—ฎ ๐—ณ๐—น๐—ผ๐˜„๐˜€ enabling one-way transfer of classified intelligence to ๐—–๐—ผ๐—บ๐—บ๐—ผ๐—ป ๐—ข๐—ฝ๐—ฒ๐—ฟ๐—ฎ๐˜๐—ถ๐—ผ๐—ป๐—ฎ๐—น ๐—ฃ๐—ถ๐—ฐ๐˜๐˜‚๐—ฟ๐—ฒ (๐—–๐—ข๐—ฃ) ๐˜€๐˜†๐˜€๐˜๐—ฒ๐—บ๐˜€ while maintaining ๐——๐—ข๐—— ๐˜€๐—ฒ๐—ฐ๐˜‚๐—ฟ๐—ถ๐˜๐˜† ๐—ฐ๐—ผ๐—บ๐—ฝ๐—น๐—ถ๐—ฎ๐—ป๐—ฐ๐—ฒ. ๐—ข๐—ฝ๐˜๐—ถ๐—บ๐—ถ๐˜‡๐—ฒ๐—ฑ ๐—บ๐—ผ๐—ฏ๐—ถ๐—น๐—ฒ ๐— ๐—Ÿ ๐—ฝ๐—ฒ๐—ฟ๐—ณ๐—ผ๐—ฟ๐—บ๐—ฎ๐—ป๐—ฐ๐—ฒ for field-deployed Android devices, achieving sub-2-second ๐—ข๐—–๐—ฅ ๐—ฝ๐—ฟ๐—ผ๐—ฐ๐—ฒ๐˜€๐˜€๐—ถ๐—ป๐—ด for combat-ready intelligence gathering. ๐—–๐—ข๐— ๐—˜๐—ง - ๐—œ๐—ป๐˜๐—ฒ๐—น๐—น๐—ถ๐—ด๐—ฒ๐—ป๐—ฐ๐—ฒ ๐—ฅ๐—ฒ๐—พ๐˜‚๐—ถ๐—ฟ๐—ฒ๐—บ๐—ฒ๐—ป๐˜๐˜€ ๐— ๐—ฎ๐—ป๐—ฎ๐—ด๐—ฒ๐—บ๐—ฒ๐—ป๐˜ ๐—ฃ๐—น๐—ฎ๐˜๐—ณ๐—ผ๐—ฟ๐—บ ๐—•๐˜‚๐—ถ๐—น๐˜ ๐—บ๐—ถ๐—ฐ๐—ฟ๐—ผ๐˜€๐—ฒ๐—ฟ๐˜ƒ๐—ถ๐—ฐ๐—ฒ๐˜€-๐—ฏ๐—ฎ๐˜€๐—ฒ๐—ฑ ๐—ถ๐—ป๐˜๐—ฒ๐—น๐—น๐—ถ๐—ด๐—ฒ๐—ป๐—ฐ๐—ฒ ๐—ฝ๐—น๐—ฎ๐˜๐—ณ๐—ผ๐—ฟ๐—บ serving 500+ analysts across multiple security classifications with 99.9% uptime SLA. ๐——๐—ฒ๐˜ƒ๐—ฒ๐—น๐—ผ๐—ฝ๐—ฒ๐—ฑ ๐˜€๐—ฒ๐—บ๐—ถ-๐—ฎ๐˜‚๐˜๐—ผ๐—บ๐—ฎ๐˜๐—ฒ๐—ฑ ๐—ฐ๐—ผ๐—น๐—น๐—ฒ๐—ฐ๐˜๐—ถ๐—ผ๐—ป ๐—ฎ๐—น๐—ด๐—ผ๐—ฟ๐—ถ๐˜๐—ต๐—บ๐˜€ reducing intelligence validation cycles by 70% through ML-powered requirement classification and routing.