Senior ML Scientist (Affective Computing & Biometric ML)
Reports to:
Chief Scientific Officer (Dr. Kaushik Ram)
Location:
Australia (Preferred Sydney/Melbourne; Remote welcome)
Employment:
Full-time, Senior
Compensation:
AUD $117,000 – $140,000 + ESOP
About inTruth Technologies
inTruth Technologies is pioneering a new era of emotional health by making emotion measurable as real-time biometric data—objectively, at scale. Using data from wearables, our proprietary Emotion Language Model (ELM) decodes the body's physiological signals into accurate, clinical-grade emotional insights.
We are building a new category — Emotion Biotech — that combines the precision of a clinical device, the scalability of AI, and the ethics of user-owned infrastructure. With sovereignty-by-design at our core, we give individuals and organizations full control of their emotional data, creating technology that serves, not exploits.
We're not just launching a product; we're building the infrastructure for a future where emotional intelligence is as fundamental as literacy, empowering humanity with empathy, resilience, and connection at scale.
Role Summary
Lead the design, validation, and productisation of scientifically-defensible emotion metrics from multimodal wearables and sensors—bridging neuroscience, clinical research, and production ML—to deliver validated, regulatory-ready outputs for research, clinical, and B2B deployment.
Core Responsibilities
A. Hypothesis Design & Experimental Leadership
* Study design, stimuli selection and ground-truth strategy for Phase-1 (Emotion detection) and Phase-2 (Emotion Score/predictive modelling) claims.
* Prepare pre-registered analysis plans (primary/secondary endpoints), power calculations, effect size, stratified subgroup analyses for university-led research and clinical trials.
* Produce statistical analysis plans (SAPs) that include mixed-effects models, equivalence/non-inferiority testing, calibration metrics, confidence intervals, and control for multiple comparisons.
B. Signal Processing & Multimodal Data Engineering
* Build and maintain canonical pipelines: PPG (time/frequency/nonlinear) + EDA decomposition, respiration proxies, eye tracking and video-derived facial features.
* Design artefact detection/repair (motion masks, contact quality, optical interference), resampling & time-synchronisation across devices, and per-window QC flags.
* Design and maintain the feature store and provenance metadata (unit tests & sample vectors).
C. Metric & Model Development (theory → math → code)
* Convert neuroscience constructs (valence/arousal, emotion detection, Emotion Score) into explicit mathematical definitions, weighting schemes, and testable formulas (with unit tests).
* Develop and evaluate models from classical (Random Forest, SVM, XGBoost) to sequence/probabilistic (temporal CNN/RNN/Transformer, Bayesian, mixture models) with explicit calibration, uncertainty quantification (CI, Brier/ECE), and blending diagnostics for mixed emotions.
* Implement LOSO/LOTO and nested CV protocols; report per-subgroup metrics (accuracy, sensitivity, specificity, F1, calibration).
D. Validation, Benchmark & Cross-Device Generalisability
* Execute validation matrix aligning datasets (lab EEG+ECG+PPG, Biostrap, Garmin) to quantify ± tolerance thresholds and implement domain adaptation/cross-device calibration.
* Produce equivalence tests for cultural and demographic cohorts and robustness tests across physiological, lifestyle and medical confounds.
E. Emotion Score and Longitudinal/Predictive Modelling
* Implement and validate Emotion Score (spike detection, α/β weighting, sampling windows, scaling/clipping).
* Build longitudinal models for mood trajectories and early-warning signals (clinical risk flags), with mixed models and survival-type analyses as required.
F. Productisation & MLOps
* Package validated metrics into production artefacts (SDKs/APIs, model registry, Docker images), define CI/CD for model training and deployment, and define telemetry/monitoring for model & data drift.
* Define endpoint SLAs for cloud and feasible on-device inference specs (quantisation/latency/energy budgets).
G. Reporting, Reproducibility & Publication
* Document all protocols, code, validation steps in Confluence/Repo; deliver whitepapers, conference abstracts and peer-reviewed manuscripts aligned to open science where permissible.
* Partner with the Board/CRO/CSO on regulatory strategy (TGA/FDA/SaMD classification) and HREC/HRA submissions.
H. Ethics & Governance
* Data minimisation, de-identification, and secure data handling consistent with HIPAA/GDPR principles and Australian HREC requirements.
* Work with in-house legal/compliance to ensure data flows meet MRFF and trial requirements (eConsent, data retention, re-consent, access control, audit logs).
Must-Have Qualifications & Experience
* PhD (preferred) or MSc + significant relevant experience in Affective Computing, Computational Neuroscience, Biomedical Engineering, Statistics, or Applied ML with demonstrable track record.
* Hands-on expertise with PPG sensors along with time-series signal processing (IBI, EDA, or ECG feature extraction and artefact mitigation).
* Proven experience designing and executing controlled experiments with physiological ground truth (PPG/EEG/clinical endpoints) and strong statistical skills (mixed models, power analysis, equivalence testing).
* Demonstrable experience in time-series ML and multi-modal fusion (sequence models/temporal CNNs/multimodal transformers).
* Production ML skills: Python (NumPy/Pandas/SciPy, scikit-learn), PyTorch/TensorFlow, XGBoost, containerisation (Docker), model versioning (MLflow/DVC) and CI/CD basics.
* Experience writing SAPs, HREC/HRA/IRB submissions, and working with clinical partners (university hospitals or lab partners).
* Exceptional written documentation skills (Confluence) and the ability to present technical results to clinical/regulatory audiences.
Desirable
* Prior experience with Biostrap, Garmin devices, Biopac, Shimmer or Lab Stream Layer integrations.
* Experience with EEG experiments and probabilistic/Bayesian modelling.
* On-device quantisation/edge deployment expertise (TF Lite/TF JS, pruning, 8/4-bit quantisation).
* Prior SaMD regulatory pathway involvement (TGA/FDA) or ISO 13485 familiarity.
* Demonstrated peer-reviewed publications in affective computing, psychophysiology, or related domains.
Success Metrics (KPIs)
The Senior ML Scientist will be evaluated against:
1. Scientific Rigor
– Delivery of at least two pre-registered analysis plans and SAPs per trial with peer-reviewed statistical validity.
2. Pipeline Robustness
– Development of multimodal pipelines (PPG + EDA + additional sensors) with ≥95% artefact detection accuracy within first 6 months.
3. Model Performance
– Emotion detection models achieving ≥70% balanced accuracy on benchmark datasets, with subgroup calibration metrics reported.
4. Emotion Score Delivery
– Prototype Emotion Score (0–100 scale) delivered within 6 months; validated across ≥2 datasets within 12 months.
5. Cross-Device Validation
– Demonstrated generalisability between at least two wearable platforms (e.g., Garmin & Biostrap) with tolerance thresholds defined.
6. Productisation
– Packaging of validated models into SDK/API artefacts with CI/CD pipeline established within first year.
7. Documentation & Publication
– Confluence/Repos fully up-to-date, plus at least one whitepaper or conference submission within first 12 months.
8. Compliance & Ethics
– Zero lapses in GDPR/HIPAA/HREC compliance across all experiments and data pipelines.
Probation & Review Clause
This role is subject to a structured performance review after the initial 90-day period. The review will assess delivery of defined milestones, scientific reproducibility, collaboration with cross-functional teams, and alignment with leadership expectations. Continuation in this role is contingent upon satisfactory performance during this period.
How to Apply
Please send your
resume and a cover letter
outlining your experience and motivation for this role to
. Applications will be reviewed on a rolling basis.