Senior ML Engineer

We’re Devopsbay - MLOPS, DevOps and AI Specialists. We know how nodes works, how to make the cloud cheaper or adapt AI to boost any area that companies need (any many more). We support our clients with strong engineers on a project basis and are always on the lookout for stellar performers. Our clients are at the cutting edge of modern solutions. We also develop our inhouse products: https://descrb.com/ & https://defencebay.com/ 

Our client is building a mobile app that uses computer vision to identify fish species, classify sex, and estimate length from photos. We’ve proven the concept — species classification at 87.6% accuracy, a length regression model with 73% improvement over baseline, and CoreML export running on iOS. Now we need someone to take this from working prototype to production-grade ML system. The founder has built the initial models and pipeline in Python, and your job is to improve model accuracy, build the infrastructure to support continuous improvement, and extend deployment to Android. This is a hands-on, build-it-yourself role — not a management position.

What You’ll Own

Design and build a repeatable, scalable process for adding new fish species to the classifier. Today, adding a species involves manual image scraping, hand-curated CSVs, and ad hoc retraining. We need a system where a new species can go from “we want to support this” to “it’s live on device” with minimal manual effort.

What We’re Looking For

5+ years of experience in ML engineering, with meaningful time spent on both model development and production infrastructure.
Specifically:

• Strong PyTorch experience, ideally with vision transformers (ViT/timm) and object detection (YOLO/ultralytics)

• Familiarity with the full model lifecycle — training, evaluation, export, and on-device deployment

• Hands-on experience deploying models to mobile (CoreML and/or TFLite); you don’t need to be an iOS/Android app developer, but you need to produce optimized model artifacts that mobile engineers can integrate

• Experience building ML pipelines with proper versioning, experiment tracking, and CI/CD

• Comfort working with small, noisy datasets and knowing how to get the most out of limited labeled data

• Experience designing classification systems that scale — adding new classes without full retraining or accuracy regression

• Familiarity with at least one orchestration tool (Airflow, Prefect, Dagster)

• Self-directed — you’ll be working with a small team and need to make pragmatic architecture decisions without a lot of oversight

Nice to Have

• Experience with MediaPipe or similar pose/landmark detection frameworks

• Background in ecological or biological image classification

• Experience with on-device A/B testing frameworks

• Familiarity with edge inference optimization (quantization, pruning, distillation)

• Experience with few-shot learning or other techniques for bootstrapping new classes with limited data