Machine Learning Engineer

T H E  R O L E
Own the intelligence layer of Rx360 — the models that turn continuous sensor and medication data into a working understanding of each individual we serve, and turn that understanding into actions that help them stay independent. You will set the ML architecture, build the longitudinal modeling stack from raw signal through to user-facing recommendation, and define how our models and our conversational companion work together.

W H AT Y O U ' L L D O
Own the ML architecture for personalized health intelligence — per-user baselines, N-of-1 and hierarchical models, and the strategy for learning from one individual's history rather than a population mean.
Build the digital biomarker stack on raw wearable signal: gait and cadence, sit-to-stand and postural transitions, activity classification, sleep and circadian structure, and cardiovascular trend from IMU and PPG streams.
Design change detection and event-association models that link medication starts, dose adjustments, and adherence gaps to downstream behavioral shifts — with defensible handling of confounding, lag structure, and multiple comparisons.
Turn a detected change into a recommendation: what to surface, to whom (the individual, family, pharmacist, or provider), with what urgency — and when the right answer is to say nothing.
Build the data and training infrastructure that carries high-frequency time-series from device ingestion through feature generation, training, serving, and monitoring.
Own evaluation. Offline benchmarks, labeled ground truth from pilot cohorts, calibration and drift monitoring, and the false-positive budget that decides whether people trust what we tell them.
Ground our conversational companion in real signal. Our AI companion runs on a large language model inside our
HIPAA-compliant cloud environment; you will connect it to per-user model output and own the evaluation and post- training work that keeps its guidance accurate, useful, and safe.
Push inference to the edge where latency, privacy, or battery demand it, working directly with the firmware and hardware teams on what runs on the band versus in the cloud.
Direct AI coding agents to accelerate implementation while holding a rigorous quality bar — the way the rest of our engineering team works.
Partner with clinical collaborators, including our university pilot, to design studies that produce the labeled data these models actually need.
Set technical direction and raise the bar for ML across the team.

R E Q U I R E D
6+ years in machine learning, including systems you have shipped to real users and operated in production — not only research or prototypes.
Deep experience with time-series or sensor data: physiological signals, IMU / accelerometry, or comparable high- frequency multivariate streams.
Demonstrated work in individual-level modeling — per-subject baselines, hierarchical or Bayesian methods, anomaly and change-point detection, or models that adapt as data from a single person accumulates.
Strong applied statistical judgment. You can separate a real effect from a coincidence in observational data, and you know exactly where a causal claim stops being defensible.
Production-grade Python and PyTorch (or equivalent). You write software other engineers are willing to maintain.
Ownership of the full lifecycle: data pipelines, training infrastructure, deployment, monitoring, and retraining.
Experience with large-language-model evaluation and post-training — grounding and retrieval, structured output, eval harnesses, and guardrails.
Cloud-native ML at production scale (AWS preferred).
Fluent use of AI development tools (Claude Code, Cursor, Copilot, or equivalent) with the domain depth to catch where generated code fails silently.
Clear, direct communication. You can set direction across firmware, backend, mobile, design, and clinical stakeholders.

N I C E T O H AV E
Wearables, digital biomarkers, gait analysis, fall risk, or human activity recognition.
Self-supervised or foundation-model approaches for wearable time series, and comfort working under real label scarcity.
Causal inference on observational data — interrupted time series, difference-in-differences, or synthetic control.
Medication-effect modeling: polypharmacy, adherence, or PK/PD intuition.
On-device and edge ML — quantization, distillation, TinyML, inference under a power budget.
Regulated environments: HIPAA and PHI handling, model governance, or the general-wellness boundary.
Clinical research collaboration and IRB-governed data.
Voice or audio ML.

W H Y T H I S R O L E I S D I F F E R E N T
G R E E N F I E L D
You are defining the intelligence layer, not maintaining someone
else's. The architecture decisions are yours to make.
R E A L S I G N A L
Continuous, longitudinal, multi- stream data from a device we build
ourselves — with the firmware team a message away.
I T M AT T E R S
A model that catches a medication effect three days early changes
whether someone keeps living in their own home.

H O W W E W O R K
Mission comes first.
The work has real impact on people's lives. We take that
responsibility seriously.
Thoughtful technology.
Technology should simplify life, not complicate it.
Respect for independence.
The people we build for value autonomy and dignity.
Everything we create reflects that.
Curiosity.
We encourage questions, experimentation, and new ways
of thinking.

A B O U T R X 3 6 0
Rx360 is a wellness platform for older adults who want to keep living on their own terms — and for the families, pharmacists, and providers around them.
The wearable is a wrist-worn band we design in-house, carrying a PPG sensor for cardiovascular signal, a six-axis IMU for motion, gait, and posture, and on-device fall detection. The iOS and Android apps give the individual a plain- language view of their own day and their medications, and open a shared circle of care to the family members they choose to include.
Rx360 is a U.S. company headquartered in Los Angeles, California, with a distributed engineering team. Hardware, firmware, mobile, backend, and machine learning are all in-house.