● ● ●

Ongoing projects 2025-present

01 LATENT HEAT

Latent Heat is a subjective archive of the communal energy that gathers in music subculture spaces. It records what photography and video cannot: the collective bodily output of a crowd, the heat, sweat, smoke, breath and movement that a room full of moving bodies generates across a night.

The intensity of a crowd is real, shared and almost entirely unrecorded. Latent Heat captures it as environmental data and gives it back as sound.
A roving archiver logs these atmospheric traces live at real events, and the installation replays that archive as a generative physical soundscape, performed by a field of autonomous electromechanical instruments, each translating one stream of data into a distinct mechanical sound.

The work treats the dancing crowd as a permeable interface between body and environment, and the gathering as something worth archiving on its own terms: speculative, decentralised, communal, and felt rather than filmed.
The archive is not a recording in the usual sense but a record of intensity, the shape of an evening, how the energy of a space rose, held and fell, rendered as something you stand inside rather than something you watch.

Discipline: data sonification · physical computing · installation

Sensing: temperature and humidity (DHT22), carbon monoxide (MQ-7), ethanol (MQ-3)

Output: five autonomous electromechanical instruments and a generative soundscape

Stack: Raspberry Pi · Arduino · Ableton

Role: concept and research, system architecture, sensor and firmware integration, the data-to-sound mapping, the Ableton signal chain, and the industrial design and fabrication of the instruments

Diagrams and maps

Process and development of instruments using servo motors, springs and kinetic repetition through Arduino code.

Exhibition and sound testing.

SWEAT-TO-SCENT

Sweat-to-Scent Interface is a research-led olfactory machine that translates athletic exertion into scent. A wearable patch collects sweat while sensors read heat, humidity, movement and conductivity. These signals are interpreted as shifting body states, and a constrained AI composes a scent recipe that a transparent four-pump machine physically mixes. Rather than reducing the body to perfomance metrics, it makes a volatile portrait of effort - a somatic archive that can be smelled, carried and remembered.

Work in Progress
2026 - present

Currently in collaboration with the MIT Media Lab Responsive Environments Group.

Click press to play

01 · The idea Not a fitness tracker, not a medical device. A somatic olfactory machine that treats sweat, heat and movement as a bodily archive — translating exertion into an olfactory language.

02 · How the system flows Body in exertion → wearable sweat patch → sensors (heat, humidity, movement, conductivity) → baseline-normalised signals → body-state model → constrained AI composes a recipe → four pumps dispense diluted oils → a scent vial / card to take away.

03 · The wearable patch A sweat-collection patch gathers bodily residue: a mesh skin layer and absorbent pad wick sweat into a small reservoir, with a dry pocket for the electronics. Built to keep sweat sealed on the body side and the sensing reliable.

04 · Sensing the body Heat and humidity (BME280), movement intensity (LSM6DS3 IMU), skin/pad temperature (DS18B20 probe) and sweat conductivity (DFRobot TDS via an ADS1115 ADC). The sensors don't read sweat perfectly — they offer interpretive signals, read against personal baselines rather than universal thresholds.

05 · Reading exertion (not diagnosing it) A deterministic layer first classifies the readings into six body states — quiet baseline, warm-up, sustained endurance, thermal strain, recovery, or ambiguous. When signals conflict, the system stays with ambiguous rather than forcing a confident answer.

06 · AI as scent composer The model isn't the scientist — it's a constrained composer. Within strict rules (only the four scents, percentages summing to 100, careful language, no medical claims) it returns a recipe as JSON, with a confidence and a one-line rationale.

07 · The machine A transparent cast-acrylic enclosure (240 × 320 × 140 mm) holds four scent reservoirs above a pump shelf, with the electronics bay and a central mixing vial. Made transparent on purpose — the process is meant to be watched.

08 · The four scents Each oil marks a quality of effort: green tea — cooling, recovery, return to baseline; cedar — endurance, rhythm, structure; mineral / ozone — sweat, salt, the electrical body; smoke / oud — strain, friction, the fatigue edge. Oils are diluted in DPG (~20%).

09 · Pumps and electronics A Seeed XIAO ESP32S3 drives a four-channel MOSFET board and four 12V peristaltic pumps, with an OLED, rotary encoder and button for control. Pumps are timed to the composed recipe.

10 · The live visual A somatic chromatogram in TouchDesigner: capillary channels, heat bloom, salt and mineral traces, pump-activation pulses and scent-vector columns — driven live over OSC by the same sensor and recipe data.

11 · The output A scent vial or printed card: the body state, the recipe, a short trace, a time. A volatile data object — an exertion portrait you can carry and re-smell.

Arduino IDE code for working prototype system using an ESP32-S3.

Working prototype integrating OLED interface, rotary encoder, MOSFET driver and peristaltic pump.

Close up of scripted code showing on the OLED screen. Showcasing the Sweat-to-Scent menu, options and process of mixing.

Sensor readings are normalised against personal baselines and classified deterministically into one of six body states — warm-up, endurance, thermal strain, recovery, quiet baseline, or ambiguous.

The code holds the logic, not the interpretation: when signals conflict it returns ambiguous rather than forcing a result.

The body state is passed to a constrained language model acting as a scent composer. It works within strict rules — only the four available scents, percentages summing to 100, no medical claims — and returns a recipe as JSON.

Each output carries a confidence and a one-line rationale in careful language — 'anchors,' 'marks,' 'traces' — keeping the system poetic but disciplined.

03 KINETIC ECHO: KE-01

Kinetic Echo is a handheld instrument that reads how you move and returns it
as image, sound and touch.

Rather than counting steps or scoring performance, it reads the texture of
movement, its flow, its bursts, its tension, and composes with it.

It is a small argument that the moving body is a creative instrument, not a
set of metrics to optimise.

Year: 2026, ongoing

Role: Concept, hardware, firmware, interaction design, sound design

Type: Handheld instrument, physical computing, real-time audiovisual and
haptic system

Hardware: Raspberry Pi Pico 2 (RP2350), GC9A01 1.28in round display,
MPU6050 six-axis motion sensor, three tactile buttons, piezo buzzer,
vibration motor

Software: C++ firmware (Arduino, Earle Philhower Pico core) computing the
six-quality pipeline on device, Python serial-to-MIDI bridge, Ableton Live

Themes: Movement as creative medium, the body as interface, feminist sport
theory

Status: Working prototype on breadboard, custom carrier PCB in development

GERBER FILES FOR CUSTOM PCB BOARD

CARRIER PCB, FIRST REVISION
Ordered as a route to final assembly. Hand-soldering proved impractical at
this pad geometry and the board was not populated.

HOW IT WORKS:

1 - It reads six qualities of movement

An inertial sensor inside the device tracks acceleration and rotation many times a second. From that raw motion the firmware continuously extracts six qualities:

Flow: how smooth and continuous the movement is
Burst: sudden accelerations, spikes of energy
Tension: held, resisted or restrained motion
Rhythm: regularity and repetition
Drift: slow, gradual change
Recovery: the settling and return to stillness

Each quality is a value that rises and falls with how you move. Together they form a signature: a fingerprint of a gesture rather than a score out of ten.

2 - The screen shows a live feed

The round display renders a field of characters, black on white, divided
into six wedges, one for each quality. As you move, each wedge fills with
denser characters as its quality rises and thins as it falls, so the screen
becomes a live portrait of your movement. Density is highest at the centre
and falls away toward the rim. Held still, the field empties to near blank.

3 - Capture freezes the signature

A press of the capture button freezes the current bloom. The live portrait becomes a held signature, a single captured moment of movement locked on the screen. This is the hold in capture, hold, replay: the gesture is no longer passing through, it is kept.

4 - Replay plays it back through the body

Pressing replay returns the held signature to the hand as a single pulse of
buzzer and vibration motor together. Its length is set by the signature
itself, weighted toward burst, then tension and flow, so a more energetic
gesture returns as a longer pulse and a quiet one as a short tap. The
signature you saw on the screen becomes something you feel.

5 - Release returns to live

Releasing clears the held signature and returns the device to live mode, following your movement again, ready for the next gesture. Capture, replay and release form the whole loop, and every gesture cycles through it.

6 - The same movement shapes sound

Alongside the screen, the six values stream out of the device over USB into Ableton Live through a small Python bridge. Each quality is mapped to a musical parameter:

Flow to filter cutoff
Burst to note triggers and accent
Tension to reverb and space
Rhythm to delay
Drift to modulation
Recovery to volume and rest

The result is that the same movement that lights the screen also shapes the sound. Motion becomes instrument.

CODE SCRIPTS IN ARDUINO AND THONNY

MAP TO ABLETON FROM DEVICE AND TERMINAL

RENDERED PROTOYPE VARIANTS

04 GLYPH UNBOUND

Nothing's Glyph language lives on the back of the phone.
Glyph Unbound takes that same visual grammar and gives it a body. It is a small round object, built to be held and worn, that renders the identical 25 by 25 dot format the Glyph Matrix uses, so a glyph made for one is a glyph the other could show.

The object has two ways in. You handle it: flip it face down and it falls to a single breathing dot for do not disturb, shake it and it rolls a Magic 8-ball answer, let it rest and it breathes quietly.
Or your phone speaks to it: a notification arrives as a pulse, and a glyph you compose on the phone appears on the screen.
Every change is confirmed by a short tone.

Alongside the object sits a small set of browser tools that send glyphs to the screen over Web Bluetooth.
You can draw a glyph on a grid, draw one in the air with your hand through live tracking, or turn your face into a field of dots and send that.
On the phone they come together as a single installable app, so the whole thing behaves like a product rather than a demo.

Glyph Unbound is an argument that a design language can be extended, not just applied.
It takes Nothing's own grammar and asks what it becomes when it leaves the phone and turns into an object you carry, hold, and read at a glance.

Year: 2026

Role: Concept, hardware, firmware, software, enclosure design

Type: Standalone object, physical computing, dot-matrix light interface, companion web tools

Hardware: ESP32-S3 round-display board, 412 by 412 dot-matrix screen, QMI8658 motion sensor, speaker, LiPo battery, 3D-printed enclosure

Software: Arduino and C++ firmware, Web Bluetooth companion tools, live hand and face tracking, installable web app

Themes: Extending Nothing's Glyph language to its own object, the body as interface, ambient computing

Status: Working prototype, with companion app and public code repository

HOW IT WORKS:

A round dot-matrix screen renders a 25 by 25 Glyph language, the same format as Nothing's Glyph Matrix.

Physical handling drives it: flip face down for do not disturb, shake for the Magic 8-ball, rest for the idle breathing state.

A phone sends notifications and glyphs to the object over Bluetooth.

Companion browser tools let you compose a glyph, draw one in the air with hand tracking, or turn your face into dots, and send it straight to the screen.

PROCESS AND DEVELOPMENT:

IMPERIAL COLLEGE LONDON

UPCOMING PROJECT:

INTERFACE WITH ELECTRONIC SKIN FOR PERSONALISED HUMAN-ROBOT INTERACTION

Versatile robotic assitance requires accurate inference of user intent. Tactile e-skins offer a direct modality for capturing high-resolution grasp data. Mapping force distributions to user posture and movement intention provides a scalable pathway towards estimating the user’s arm’s force and rigidity, enabling intuitive and personalised collaboration with assistive robots.

This project will use an interface with electronic skin (e-skin) to develop personalised human-robot interaction. A handle mounted on a collaborative robot will be equipped with a 16x16 force-sensing e-skin wrapped around its surface. The aim is for the robot to execute a range of interaction conditions while a user graps and manipulates the handle, record the resulting spatial force and predict short-horizon movement intention from tactile data.

Year: 2026

Dates: 20th-24th July

Research Group: Human Robotics Group

Project Supervisor: Prof. Etienne Burdet

Hardware and Software stack: NunPy, PyTorch, scikit-learn, UR, Franka, Kuka, deep learning on time series/image-like sensor data, EMG, IMU and biosignal sensors, human biomechanics and taxonomy.