ML Data Pipeline Engineer

We're seeking a Data Pipeline Engineer to own and evolve our exercise recognition training data infrastructure. You'll manage the end-to-end pipeline that collects, synchronizes, validates, and prepares IMU sensor and video data for ML model training.

This role combines systems engineering, data quality automation, and hands-on problem-solving in a production environment.

What You’ll Do

Pipeline Operations & Improvement

Maintain and enhance our multi-source data collection system : IMU sensors (via mobile app) and synchronized video streams from gym-based cameras.

Improve video capture software robustness, particularly handling network interruptions and operational monitoring.

Deploy and monitor services in remote Linux environments with appropriate DevOps practices.

Data Quality & Validation

Evolve our Python-based QC engine that validates data pre- and post-annotation

Implement checks for IMU-video time synchronization, sensor health, and measurement consistency

Apply digital signal processing techniques to identify sensor failures, connectivity issues, and measurement irregularities.

Develop validation logic comparing annotations against sensor data to ensure temporal alignment.

Analysis & Troubleshooting

Perform ad-hoc analysis on ~1,200+ workout tasks to classify failure modes

Identify whether issues stem from pipeline bugs, sensor problems, or annotation errors

Prioritize engineering work based on data quality impact and coordinate with annotation team on fixes

Tooling and Visualization

Maintain and extend our NextJS UI serving annotators, data scientists, and stakeholders

Create visualizations (Chart.js) for QC metrics and signal analysis

Integrate with LabelStudio annotation interface

What You Bring

Required

Strong Python programming skills, particularly for data processing pipelines

Experience with time-series data and digital signal processing

Comfortable working in Linux environments and deploying / monitoring remote services

Ability to debug complex multi-component systems (sensors, video, networks, sync)

Data quality mindset : designing validation rules, tracking metrics, investigating anomalies

SQL / database experience for managing pipeline metadata

Highly Valued

Video processing experience (RTSP streams, encoding, OCR)

Working with sensor / IoT data and handling connectivity challenges

NextJS or modern web frameworks for data tooling

DevOps practices : containerization, monitoring, logging, alerting

Experience with annotation pipelines and ML training data workflows

Background in biomechanics, sports science, or wearable sensors

Tech Stack

Languages : Python (primary), JavaScript / TypeScript (NextJS UI)

Data : IMU sensor streams, video (RTSP), time-series analysis, DSP

Tools : LabelStudio, Chart.js, Linux / bash, OCR libraries

Infrastructure : Remote deployment, monitoring systems

You'll Thrive Here If You

Enjoy detective work : diagnosing why data doesn't match expectations

Balance pragmatism with quality : shipping improvements while maintaining reliability

Communicate well across technical and non-technical stakeholders

Can work autonomously in a small, mission-driven team