Production AI Research Engineer - OCR & Document Extraction (NYC)

Endless Labs, Inc.; New York (NY); Hybrid; USD 275,000 - 325,000; Full time

Job summary

Datalab is seeking a Research Engineer to own problems end-to-end across models, inference service, and product. You will take models from training through benchmarking and into our inference stack, collaborating with the team to integrate them into products.

You'll work on training small, efficient models that outperform larger LLMs on domain-specific tasks like OCR and structured extraction, while balancing exploration with shipping velocity in a fast-paced startup.

Qualifications

• 3+ years experience training, fine-tuning, and evaluating deep learning models.
• Trained at least one production-grade model or system used in real-world applications.
• Deep expertise in PyTorch and Python, with strong fundamentals in deep learning (optimization, evaluation, architecture design).
• Comfortable with data engineering, benchmarking, and performance profiling across hardware setups.
• Comfortable with an early stage startup - balance running ablations/benchmarks with shipping velocity.

Responsibilities

• Train and evaluate models: Train task-specific models (OCR, layout, text recognition, extraction).
• Optimize inference: Profile and accelerate model inference across hardware (H100s, B200s, L40s, CPUs).
• Ship to product: Integrate models into API and product, shaping how capabilities surface for users.
• Create and maintain datasets: Source, design, and clean datasets; build reproducible pipelines.
• Experiment and benchmark: Run ablations, track metrics, publish findings.
• Engage with users and partners: Join calls/slack threads to understand customer needs.

Skills

DL training

Model fine-tuning

Python programming

Deep learning fundamentals

Performance profiling

Benchmarking

Startup environment

Tools

PyTorch

Python

Job description

Datalab is seeking a Research Engineer to own problems end-to-end across models, inference service, and product. You will take models from training through benchmarking and into our inference stack, collaborating with the team to integrate them into products.

You'll work on training small, efficient models that outperform larger LLMs on domain-specific tasks like OCR and structured extraction, while balancing exploration with shipping velocity in a fast-paced startup.