R Scientist for Health Signal Processing
Budget: $250 – $750 USD
Freelance R Expert Needed: Advanced Signal Processing for Maternal, Fetal & Neonatal Health
We are seeking a senior R scientist with deep experience in audio / seismic data, wavelet methods, and Bayesian modeling to help demonstrate that R is just as good as Python at surfacing early physiological health markers.
This role is for someone comfortable working below the noise floor—where the signal of interest is subtle, nonstationary, and embedded in the mechanics of living systems.
What you’ll work on
1. Signal conditioning & noise removal
• Design and evaluate signal conditioning pipelines for low-amplitude physiological data
• Implement filtering strategies to remove mains contamination (50/60 Hz and harmonics), including:
o Notch / comb filters
o Adaptive and time-varying filters
o Evaluation of filtering impact on downstream features
• Distinguish true physiological oscillations from environmental and instrumentation noise
2. Time–frequency feature extraction
• Wavelet-based decomposition of vibroacoustic signals
• Work with TLSW-style objects to estimate:
o Smooth latent trends
o Time-varying wavelet spectra
• Visualization using:
o plot.default for estimated trends
o plot.wd for wavelet spectra (Guy Nason–style visualizations)
o Custom control via trend.plot.args and spec.plot.args
3. Feature importance & interpretability
• Quantify and compare time-domain vs frequency-domain feature importance
• Assess which features carry predictive power across:
o Gestational age
o Maternal stress and hemodynamics
o Fetal and neonatal state transitions
• Use interpretable frameworks to determine:
o Which frequency bands matter
o When in time those features emerge
o How feature importance shifts longitudinally
• Connect statistical importance to physiological plausibility
4. Bayesian & longitudinal modeling
• Bayesian hierarchical and state-space models
• MCMC / Bayesian Markov Chain Monte Carlo workflows
• Longitudinal and mixed-effects models for repeated measures
• Subject-specific trajectories and population-level inference
• Uncertainty-aware estimation (posterior diagnostics, credible intervals)
What we’re looking for
You are likely a strong fit if you have:
• Advanced proficiency in R for statistical and signal processing work
• Hands-on experience with raw audio / time-series data
• Strong intuition for:
o Wavelets and spectral analysis
o Filter design and signal conditioning
o Feature selection and interpretability
• Practical experience with:
o Bayesian hierarchical models
o MCMC (Stan, JAGS, custom implementations)
o Longitudinal and mixed-effects modeling
• Comfort translating math and code into biologically meaningful insight
Why this is interesting
• You’ll work on post-hoc classification, time-series longitudinal data (regression) and individuation for early-warning biology
• The problems are technically demanding and scientifically consequential
• This is a chance to apply serious signal processing where it can change outcomes—not just metrics
Engagement details
• Include the word for mother in your native tongue in your response
• Please include a brief description of your experience with signal conditioning, wavelets, Bayesian models, and feature importance analysis in R.
• Opportunity for longer-term collaboration
We’re especially interested in people who know how to use R to look for the signal hidden in the “noise”.
We are seeking a senior R scientist with deep experience in audio / seismic data, wavelet methods, and Bayesian modeling to help demonstrate that R is just as good as Python at surfacing early physiological health markers.
This role is for someone comfortable working below the noise floor—where the signal of interest is subtle, nonstationary, and embedded in the mechanics of living systems.
What you’ll work on
1. Signal conditioning & noise removal
• Design and evaluate signal conditioning pipelines for low-amplitude physiological data
• Implement filtering strategies to remove mains contamination (50/60 Hz and harmonics), including:
o Notch / comb filters
o Adaptive and time-varying filters
o Evaluation of filtering impact on downstream features
• Distinguish true physiological oscillations from environmental and instrumentation noise
2. Time–frequency feature extraction
• Wavelet-based decomposition of vibroacoustic signals
• Work with TLSW-style objects to estimate:
o Smooth latent trends
o Time-varying wavelet spectra
• Visualization using:
o plot.default for estimated trends
o plot.wd for wavelet spectra (Guy Nason–style visualizations)
o Custom control via trend.plot.args and spec.plot.args
3. Feature importance & interpretability
• Quantify and compare time-domain vs frequency-domain feature importance
• Assess which features carry predictive power across:
o Gestational age
o Maternal stress and hemodynamics
o Fetal and neonatal state transitions
• Use interpretable frameworks to determine:
o Which frequency bands matter
o When in time those features emerge
o How feature importance shifts longitudinally
• Connect statistical importance to physiological plausibility
4. Bayesian & longitudinal modeling
• Bayesian hierarchical and state-space models
• MCMC / Bayesian Markov Chain Monte Carlo workflows
• Longitudinal and mixed-effects models for repeated measures
• Subject-specific trajectories and population-level inference
• Uncertainty-aware estimation (posterior diagnostics, credible intervals)
What we’re looking for
You are likely a strong fit if you have:
• Advanced proficiency in R for statistical and signal processing work
• Hands-on experience with raw audio / time-series data
• Strong intuition for:
o Wavelets and spectral analysis
o Filter design and signal conditioning
o Feature selection and interpretability
• Practical experience with:
o Bayesian hierarchical models
o MCMC (Stan, JAGS, custom implementations)
o Longitudinal and mixed-effects modeling
• Comfort translating math and code into biologically meaningful insight
Why this is interesting
• You’ll work on post-hoc classification, time-series longitudinal data (regression) and individuation for early-warning biology
• The problems are technically demanding and scientifically consequential
• This is a chance to apply serious signal processing where it can change outcomes—not just metrics
Engagement details
• Include the word for mother in your native tongue in your response
• Please include a brief description of your experience with signal conditioning, wavelets, Bayesian models, and feature importance analysis in R.
• Opportunity for longer-term collaboration
We’re especially interested in people who know how to use R to look for the signal hidden in the “noise”.