Optical physics engineer leading the development of AI vision models and a spectral analysers for particulate pollutants | Ingénieur en physique optique, spécialisé dans le développement de modèles de vision par IA et d'analyseurs spectrales pour les polluants particulaires.

- Producer and host of the Envirohealth Podcast. - The Envirohealth Podcast explores all things environment and health from measurement to mathematics, data to diagrams, publication to policy.

- Autonomously spearheaded the creation of an online spectral database and chemometric software tailored for plastic researchers. - Developed an intuitive web-based front-end utilising the JavaScript framework React, linked to a Firebase Database. - Employed REST APIs for client-side interactions with the database, underpinning calls through Firebase's callable HTTPS requests. - Created automated signal processing pipelines that incorporate techniques like spectral alignment (standard normal variate), baseline correction (asymmetric partial least squares regression), artefact removal (eliminating cosmic rays from chemical datasets), intensity normalisation (maximum normalisation), and signal filtration (Savitzky-Golay filtration). Implemented a Python virtual environment within the Firebase Function suite and curated univariate and multivariate chemometric statistical scripts. These scripts enable the categorisation of a polymer's spectrum, distinguishing between Raman and FTIR spectra.- To enhance the efficiency of classification processes, I engineered a deep residual network, boasting an impressive composition of over 60 layers. This model underwent rigorous training on synthetically augmented Raman and FTIR spectra, expanding the database to encompass more than 10,000 spectra. The synthetic augmentation techniques introduced included the incorporation of scattering and baseline noise. Developed with PyTorch, this deep residual network demonstrates a classification accuracy of approximately 95% for virgin polymers, with limited chemical additives. However, further data is requisite for precise classifications of polymers infused with dyes or other chemical additives that significantly influence a polymer's spectrum. - This open-source software is readily accessible to spectroscopists globally, facilitating the meticulous categorisation of their spectral datasets.

- Established an educational charity based in London – encouraging young people from historically marginalised communities to forge new paths by providing curriculum and supplementary educational support that incorporates "in-person" and "online" learning experiences.- Leading a team of 40 volunteers to co-create an organisation that propel engaged students across London and globally (India and Indonesia) to improved academic attainment. Providing students an opportunity to interact with leading and budding experts on topics ranging from AI to Fine Art. - Designed, optimised, and built an online learning platform (Educo) using the JavaScript framework, React in collaboration with the Charity's Senior Web-development Officer. Designed and implemented a full-stack React application connected to a Firebase Real-time database, which supported the use of a Neural Collaborative Filtering (NCF) model recommending courses/videos to learns. - Built a flutter cross-platform application, which boasts the same functionality of the website in the an mobile application format (Release date: 1/05/2024).- Led the creation of an academic mentoring programme, working in collaboration with Elm Academy in Lambeth to allow for professionals to mentor students through their Extended Project Qualification. This a year-long mentoring programme, for which I and the Volunteering team facilitate.- Fully qualified Designated Safeguarding lead, co-managing safeguarding and child protection across the charity with Tim Siu.

- Conceptualised and executed a pilot occupational health assessment in Material Recycling Facilities. Receiving €120, 000 of funding from AXA Health. - Evaluated the health impact associated with employment in Material Recycling Facilities. Requiring the employment data analytics frameworks and machine learning to predict harm, and toxicological profiles of inhaled compounds using Graph Convolution Neural Network. - Developed and optimised one-convolutional deep residual neural networks in PyTorch for the prediction and chemical classification of spectral signals originating from plastics and other environmental airborne particulates. Models from such are now deployed on the plasticanalytics.com website, an open-sourced spectral library, enabling any spectroscopist, globally, to categorise their spectral datasets. - Mentored the development and optimisation of a gradient boost classification model, aiming to enhance the predictive accuracy of chemical assignments in two-dimensional gas chromatography-mass spectrometry datasets. Achieving its primary goal of better alignment between user-assigned and computationally defined chemical assignments, outperforming prior methods by ~90% accuracy. The model and results, which await publication, will be uploaded to an open-source online library for others to utilise.

- Led a national atmospheric deposition sampling campaign evaluating the presence of airborne micro- and nano- plastics across the United Kingdom.- Mentored MSc students identifying plastics in Environmental samples (soil, air, and water), and biological matrixes (serum samples). - Optimised methods for detecting microplastics using two-dimensional gas-chromatography, with a fully quantification method for four solubilised polymers. - Applying such method to PM10 and PM2.5samples collected in Barcelona and total suspended particulate samples in Pakistan. In these projects, I mentored visiting Ph.D. candidates to identify plastics in their airborne samples.

Assistant Head Coach, working in a large team of internationally acclaimed tennis coaches on a residential summer camp targeted at elite tennis players (7-16 yrs).