Machine Learning Engineer

Recruiting From Scratch•San Francisco, CA

39d•Onsite

About The Position

This is a rapidly growing AI technology company building machine learning systems that understand complex, unstructured documents and transform them into reliable inputs for modern AI applications. The company's technology combines computer vision, vision-language models, and large language models to process difficult real-world documents and make their information usable by downstream AI systems and enterprise applications. The machine learning organization is currently approximately eight people and is continuing to grow. The team works across the entire ML lifecycle, from research and experimentation through training, evaluation, deployment, and production monitoring. This role reports directly to the CTO and offers significant ownership over models and ML systems from research through production. Engineers work with a combination of smaller models developed internally and larger open-source models, with a strong emphasis on shipping systems that operate against real enterprise workloads. The ideal candidate is a hands-on ML engineer who has personally taken models all the way from experimentation to production. This is not a research-only or integration-focused role — candidates should be comfortable building, evaluating, deploying, and continuously improving production ML systems.

Requirements

4+ years of full-time machine learning engineering experience after completing a degree or equivalent education.
Experience training, fine-tuning, evaluating, and deploying ML models into production.
Demonstrated ownership of the complete ML lifecycle.
Experience shipping machine learning systems to real users or customers.
Strong production engineering experience alongside ML expertise.
Experience working with complex real-world data.
Experience with computer vision, VLMs, LLMs, or document understanding.
Strong Python experience.
Strong experience with PyTorch or comparable deep learning frameworks.
Hands-on experience with VLMs, computer vision, or LLMs.
Experience training and fine-tuning machine learning models.
Strong ML evaluation experience.
Experience with production model deployment.
Experience building data and ML pipelines.
Experience with OCR and document understanding.
Strong understanding of machine learning fundamentals.
Experience working with modern AI/ML infrastructure and tooling.
Familiarity with production software engineering practices, testing, debugging, and monitoring.
Demonstrated ability to take an ML model from research or experimentation through production.
Experience designing meaningful model evaluations.
Ability to identify and debug model quality issues.
Experience improving model performance through training, fine-tuning, data, architecture, or evaluation.
Strong understanding of the tradeoffs between model quality, latency, cost, and reliability.
High ownership and accountability.
Strong analytical and research instincts.
Comfortable operating in a fast-paced startup environment.
Able to clearly explain technical contributions and research findings.
Strong communication with engineers, founders, product teams, and customers.
Comfortable working independently on ambiguous ML problems.
Curious and highly motivated by difficult technical challenges.
Pragmatic about shipping production systems rather than pursuing research in isolation.
Comfortable receiving feedback and iterating quickly.
Strong ability to connect technical work to measurable product outcomes.

Nice To Haves

Recent experience in a startup or high-growth technology environment is strongly preferred.
Research experience and publications are strong pluses when paired with production engineering experience.
Experience working with multimodal or document-focused AI is particularly valuable.
Publications in computer vision, VLMs, document understanding, or related fields are a strong plus.
First- or second-author publications with meaningful citation impact are a strong signal.
GitHub contributions, research projects, or other evidence of technical depth are valued.

Responsibilities

Build and train state-of-the-art machine learning models for parsing and interpreting complex, unstructured documents.
Fine-tune and adapt models for real-world production use cases.
Deploy ML models into production systems used by enterprise customers.
Own model performance from initial experimentation through production deployment.
Improve model accuracy, reliability, latency, and overall production performance.
Work with Python and modern ML frameworks such as PyTorch.
Contribute directly to the development of computer vision, VLM, and LLM-based systems.
Investigate novel approaches to improve model performance and accuracy.
Experiment with VLMs, LLMs, computer vision models, and document understanding techniques.
Design experiments to evaluate different model architectures, training approaches, and fine-tuning strategies.
Develop methods for improving the quality and reliability of AI systems.
Build tooling and evaluation frameworks to determine whether new techniques meaningfully improve production performance.
Stay current with developments in machine learning, computer vision, document intelligence, and multimodal AI.
Build and maintain data pipelines supporting model training and evaluation.
Develop robust evaluation systems for measuring model performance.
Own the complete ML lifecycle from data preparation and experimentation through deployment.
Integrate trained models into production product workflows.
Monitor model behavior and identify opportunities for continuous improvement.
Build infrastructure and tooling that enables rapid experimentation and reliable production deployment.
Ensure models are production-ready rather than stopping at research prototypes.
Work closely with the CTO and founding team on ML strategy and technical direction.
Partner with product and engineering teams to translate customer problems into ML solutions.
Work directly with customers and real enterprise workloads to understand model performance requirements.
Help shape product direction through technical experimentation and ML insights.
Clearly communicate technical findings, experiment results, and model tradeoffs.
Take ownership of meaningful ML problems without relying on multiple layers of management or research handoff.