Research Engineer, End-to-End ML & OCR in NYC

datalab.mn d.o.o.; New York (NY); On-site; USD 275,000 - 325,000; Full time

Job summary

Datalab is seeking a Research Engineer to own end‑to‑end problems across models, inference service, and product. You’ll train, benchmark, and integrate models into our API and product, with emphasis on the model side and practical results.

You’ll work on small, efficient models for domain tasks like OCR and structured extraction, balancing exploration with shipping velocity in a startup environment.

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.
• 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 task‑specific models (OCR, layout, text recognition, extraction).
• Optimise model inference across hardware (H100s, B200s, L40s, CPUs).
• Ship models into our API and product, defining how new capabilities surface for end users.
• Create and maintain datasets for supervised and synthetic training; build reproducible pipelines.
• Run ablations, track metrics, and publish findings to inform design directions.
• Engage with users and partners to understand customer needs.

Skills

PyTorch

Python

Deep Learning

Data Engineering

Benchmarking

Performance Profiling

Tools

Chandra

Marker

Surya

Job description

Datalab is seeking a Research Engineer to own end‑to‑end problems across models, inference service, and product. You’ll train, benchmark, and integrate models into our API and product, with emphasis on the model side and practical results.

You’ll work on small, efficient models for domain tasks like OCR and structured extraction, balancing exploration with shipping velocity in a startup environment.