AI Time-Series Data Synthesis Engineer

Telvion Labs

Remote · $100k – $130k / year · Full-time · Posted 4 months ago

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

About the role

Tabular data synthesis is a solved problem — imperfectly, but the techniques are mature. Time-series synthesis is much harder: temporal structure, autocorrelation, seasonality, trend components, and multivariate co-movement patterns all need to be preserved for the synthetic data to be useful for model training. A synthetic financial time series that gets daily returns right but destroys the volatility clustering that models need to learn is useless for market risk model training. We develop synthetic data capabilities for 35 ML research organizations and enterprise data science teams. Time-series synthesis is a gap in our current portfolio: we support tabular, image, and text synthesis. Our clients need time-series synthesis for financial modeling, sensor data augmentation, and clinical longitudinal data generation where real data is scarce or regulatory-constrained. You'll build the time-series synthesis system: implementing conditional generation models that preserve specified temporal dependencies, developing evaluation metrics that quantify temporal fidelity beyond standard statistical moments, and building domain adaptation that lets clients specify the temporal properties that matter for their particular time-series use case.

Fully remote.

Responsibilities

Design conditional time-series generation models preserving autocorrelation, seasonality, and multivariate co-movement
Implement temporal fidelity evaluation metrics beyond distributional statistics: spectral similarity, autocorrelation distance
Build domain adaptation enabling clients to specify target temporal properties for their synthesis use case
Develop conditional generation conditioning on user-defined trend and regime change specifications
Create multivariate synthesis preserving cross-series correlation structures
Benchmark synthesis quality against domain-specific downstream task performance on synthetic versus real data

Requirements

3+ years in ML engineering with time-series modeling, generative models, or synthetic data experience
Strong Python and PyTorch skills with sequence generation and temporal model architectures
Deep understanding of time-series statistics: autocorrelation, spectral analysis, stationarity, and regime changes
Experience evaluating generative model output quality beyond standard statistical distance metrics
Ability to work across financial, sensor, and clinical time-series domains with different temporal fidelity requirements