Daniele Re, PhD

Signal processing, machine learning & AI engineering for physiological and neural sensing

Algorithm scientist & AI engineer · wearable & neural sensing

ML Consultant, X-trodes · Co-founder, Telephos Intelligence

From modelling binocular vision for a humanoid robot to decoding eye movements from electrodes on the face, my path has followed one question: how do you turn a weak, noisy signal from the body into a measurement you can trust?

I trained as a biomedical and neuro-engineer at the University of Genoa, then did a PhD in cognitive neuroscience at the Hebrew University of Jerusalem, showing that attention samples the visual world rhythmically.

As a postdoc at Tel Aviv University I built real-time closed-loop systems on wearable EMG and accelerometer streams and planned non-invasive retinal stimulation. Since then I've worked on wearable biosignal ML as a consultant to X-trodes and co-founded Telephos, where I built a multimodal capture platform for health measurement and the agentic AI service that analyses its data, from the FastAPI backend to deployment, plus a voice check-in agent and a real-time AI avatar.

Across all of it the loop is the same: design how the data is collected, understand the noise, build the model, and validate it against something trustworthy, including reporting what doesn't work.

10Publications
6Signal modalities
9+Years research
2Fellowships
Hebrew University of Jerusalem Tel Aviv University X-trodes Telephos Intelligence King's College London University of Genoa Jerusalem Brain Community Mandel School, Hebrew University InnoNext talent grant European Innovation Council
Download CV Get in touch

01Selected work

Telephos Intelligence · co-founder · 2025–now

An agentic AI service for clinical-trial data

I built and deployed the platform end to end: a FastAPI backend with authentication, audit logging and Postgres behind a React front end, and an LLM layer on the Claude API. A tool-calling assistant fetches only the data windows it needs and streams its reasoning live; other features turn plain language into charts, map incoming datasets onto our schema, and flag data-quality problems for human review.

Checked, not assumed: a gold-set evaluation scored by an LLM judge, tracing and cost tracking, versioned prompts with rollback, and 549 automated tests. Deployed with Docker.

Claude APItool callingFastAPIReactevalsDocker
Telephos Intelligence · co-founder · 2026

A real-time AI avatar and voice check-in

An evening spoken check-in: ElevenLabs turns speech into text and back, but every word is decided by our own state machine. Questionnaire items are read word for word, Claude writes only the conversational lines, and each line must pass a rule-based guard; crisis handling is fixed rules, never a model. On top, a live video avatar: a talking face rendered on a Modal GPU, streamed over LiveKit, that nods, smiles and mirrors through a versioned reaction policy, while face tracking runs in the browser so the camera video never leaves the device.

Tested like software: a simulated-patient test bench that caught real bugs, unit-tested reaction rules, measured avatar delay (about 1.3 s) and head motion benchmarked against a commercial avatar. Works on phones.

voice agentElevenLabsLiveKitModal GPUMediaPipeClaude API
X-trodes · consulting · 2024–now

Eye movements from electro-oculography on dry-electrode wearables

PyTorch transformer models that recover eye movements from EOG recorded in everyday conditions, plus the pipelines that make those recordings usable: filtering, artefact rejection, stream alignment, drift correction, signal-quality scoring.

Validated against research-grade reference instruments, with failure regimes mapped under motion, electrode noise, dropped samples and device-to-device variation.

PyTorchEOGEEGEMGIMU
Telephos Intelligence · co-founder · 2025–now

A multimodal capture platform for health measurement

The capture stack from sensor selection up: Android and Wear OS apps recording facial video, voice and smartwatch physiology, device timing and synchronisation, and the backend and pipelines behind them. It recorded a UK clinic pilot and supports ongoing US studies.

Decision: a simulation showed one planned feature couldn't be separated from reporting bias, so we dropped it and redesigned the measurement.

KotlinWear OSFastAPIDatabricks
Tel Aviv University · postdoc · 2023–2025

Real-time closed-loop sensing and non-invasive retinal stimulation

A real-time system that detected movement events from wearable EMG and accelerometer streams and triggered an intervention within the movement window. Camera pipelines for gaze, vergence, microsaccades, pupillometry and blinks. SimNIBS electric-field simulations to plan transorbital tACS and temporal-interference stimulation.

Outcome: stimulation plans used in real sessions with participants; first-author paper at IEEE EMBC 2025.

closed loopEMGeye trackingSimNIBS
Hebrew University · PhD · 2017–2023

Attention samples the world rhythmically

Does attention process several objects at once, or sample them in alternation? Dense behavioural time courses, spectral analysis, and careful statistics for periodicity claims, including a published reply to a methodological critique.

Published in Current Biology (2019) and Trends in Cognitive Sciences (2026).

psychophysicsspectral analysisstatistics

02Skills

Signal processing

Filtering, ICA and artefact rejection, spectral and time-frequency analysis, drift correction, event detection, synchronising asynchronous streams (LSL). MNE-Python, SciPy.

Machine learning

PyTorch transformer and sequence models for physiological time series, transfer from pretrained encoders, Kalman filtering and state-space models, Bayesian models in Stan, classical ML for small samples.

Validation & experiments

Benchmarks against reference instruments, subject-level cross-validation, ablations, calibration, bootstrap, mixed-effects models, power analysis, psychophysics and human-study design.

Capture systems

Multi-sensor capture with wearable electrodes, IMU, cameras and smartwatches; Android/Wear OS apps (Kotlin, Camera2, BLE); real-time closed-loop systems.

AI engineering

LLM applications on the Claude API: tool calling and agent loops, structured outputs (Pydantic), evaluation with gold sets and LLM judges, tracing and cost control. FastAPI, React and Next.js, Postgres, Docker, cloud deployment, pytest.

Field expertise

Self-assessment, 10 = where I'm strongest.

Signals & sensing

EOG & eye tracking
9/10
EEG / EMG
8/10
IMU & motion
7/10
Sensor fusion
7/10
ECG / PPG / EDA
6/10

Machine learning

Validation design
9/10
Sequence models
8/10
State-space / Kalman
8/10
Computer vision
6/10
MLOps
3/10

Neuroscience

Visual attention
10/10
Psychophysics
10/10
Oculomotor control
8/10
Neuromodulation
7/10
Binocular vision & rivalry
9/10
Neural oscillations
8/10

Engineering

Python
10/10
MATLAB
8/10
Kotlin / Android
6/10
Docker / CI
6/10
TypeScript / React
6/10
SQL / Postgres
6/10
C++
5/10

AI engineering

LLM apps & agents
8/10
LLM evals & guardrails
8/10
Backend APIs (FastAPI)
8/10
Voice & real-time AI
7/10
Cloud & deployment
7/10

Statistics & modelling

Hypothesis testing
9/10
Mixed-effects models
9/10
Spectral statistics
9/10
Bayesian models (Stan)
8/10
Simulation studies
8/10

03Timeline

Education Research & work Teaching Publication
2026
Publication

Attentional sampling resolves competition along the visual hierarchy

Trends in Cognitive Sciences · first author
Publication

Multimodal quantification of cognitive load with a printed wearable facial bio-potential system

Journal of Neural Engineering
2025
Research

CTO / CSO & Co-founder

Telephos Intelligence
2025 – present

Multimodal capture platform for health measurement: capture apps, sync, pipelines, validation.

Publication

Facial dry-printed electrodes for transocular stimulation

IEEE EMBC · first author
Publication

Dry-printed electrodes for transcutaneous stimulation of innervated muscles

J. NeuroEngineering and Rehabilitation
Publication

Sampling without vision: auditory performance fluctuates in congenitally blind individuals

Journal of Cognitive Neuroscience
Publication

Evaluation of dry printed electrodes for safe transcutaneous stimulation

IEEE EMBC
2024
Research

ML Consultant

X-trodes, Herzliya
2024 – present

Wearable biosignal ML: EOG-to-gaze transformer models, signal pipelines, reference-instrument validation.

2023
Research

Postdoctoral Fellow

Tel Aviv University
2023 – 2025

Real-time closed-loop wearable sensing; retinal stimulation planning with SimNIBS.

Education

PhD, Cognitive Neuroscience

Hebrew University of Jerusalem
2017 – 2023

JBC Doctoral Gold Fellowship and Mandel Presidential Scholarship (2019).

Publication

Attentional sampling between eye channels

Journal of Cognitive Neuroscience · first author
2022
Preprint

Perplexity about periodicity repeats perpetually

bioRxiv
2019
Publication

Feature-based attention samples stimuli rhythmically

Current Biology · first author
2017
Teaching

Lecturer

University of Genoa
2017 – 2018

Neural coding and natural scene statistics.

Publication

Binocular learned receptive fields for saccade-vergence control

VISAPP · first author
2015
Research

Research Associate

King's College London
2015 – 2016

Bio-inspired binocular vision models in the iCub robot simulator.

2010
Education

BSc Biomedical Engineering · MSc Neuro-Engineering

University of Genoa
2010 – 2016

04Publications

2026
Multimodal quantification of cognitive load using a printed wearable facial bio-potential systemTeitelbaum, D., Ibrahim, R., Man, H., Luxembourg, N., Re, D., & Hanein, Y. · Journal of Neural Engineering
2026
Attentional sampling resolves competition along the visual hierarchyRe, D., Kusnir, F., & Landau, A. N. · Trends in Cognitive Sciences
2025
Dry-printed electrodes for transcutaneous electrical stimulation of innervated musclesIbrahim, R., Vėbraitė, I., Re, D., et al. · J. NeuroEngineering and Rehabilitation
2025
Sampling without vision: auditory performance fluctuates in congenitally blind individualsKhasis, E., Re, D., Kusnir, F., Bhat, A., & Landau, A. N. · Journal of Cognitive Neuroscience
2025
Efficacy and usability of facial dry-printed electrodes for transocular stimulationRe, D., Ibrahim, R., & Hanein, Y. · IEEE EMBC
2025
Evaluation of dry printed electrodes for effective and safe transcutaneous electrical stimulationIbrahim, R., Re, D., & Hanein, Y. · IEEE EMBC
2023
Attentional sampling between eye channelsRe, D., Karvat, G., & Landau, A. N. · Journal of Cognitive Neuroscience
2022
Perplexity about periodicity repeats perpetually: a response to BrookshireRe, D., Tosato, T., Fries, P., & Landau, A. N. · bioRxiv
2019
Feature-based attention samples stimuli rhythmicallyRe, D., Inbar, M., Richter, C. G., & Landau, A. N. · Current Biology
2017
An integrated system based on binocular learned receptive fields for saccade-vergence on visually salient targetsRe, D., Gibaldi, A., Sabatini, S. P., & Spratling, M. W. · VISAPP (Int. Conf. on Computer Vision Theory and Applications)

Full list on Google Scholar →

05Contact

Building something that measures people?

I'm open to algorithm, research-scientist and ML roles in health, wearables and neurotech, in Israel, Europe or further afield. Italian (EU) citizen, based in Givatayim.

daniele.rejo@gmail.com LinkedIn Google Scholar