Rayan Daod Nathoo

Dec 2023 - Now

Junior Software Engineer at Atinary Technologies - using Machine Learning to enable self-driving labs and accelerate Research & Development by orders of magnitude (10x to 100x). Applications in Climate Tech, Energy, Life Sciences, Green Chemicals, and more. I will work at the intersection of Software Engineering and Machine Learning to build a robust and scalable platform. Exciting times ahead!

Feb - Aug 2023

ML Research Intern at Bose Corporation, supervised by Mikolaj Kegler and Marko Stamenovic to develop lightweight and low-latency Speech Enhancement and Source Separation Deep Learning models. Focus on Knowledge Distillation to improve the performance of tiny causal models subject to hard memory, energy, and latency constraints. I developed a novel method particularly good in adverse conditions (low-SNR input, sub-100k-parameter models).
> Publication (ICASSP 2024)
> Poster (SANE 2023)

May - Aug 2022

Computer Vision Engineer in the Multimedia Signal Processing Group (MMSPG) at EPFL. I worked with Prof. Dr Touradj Ebrahimi on evaluating and deploying open-source Image Segmentation and Super-Resolution Deep Learning models into a video streaming pipeline, as part of AdMiRe, an EC-funded project (Horizon 2020). The project was successfully validated by the reviewers and end users!
> Publication (SPIE 2022)
> Code

Aug 2021 - Feb 2022

Machine Learning Intern and Master's Thesis at Logitech supervised by Dr Milos Cernak and Dr Paolo Prandoni. I developed and prototyped a real-time Target Source Separator for audio event extraction/removal. The resulting Deep Learning model takes a mixture as first input, and extracts/removes target sources based on a conditioning vector given as second input. The prototypes were aimed at gamers (footstep enhancement in FPS games) and neurodivergent people (full audio environment personalisation).
> Master's Thesis
> Mention at AMLD 2022 by Dr Milos Cernak (Principal Engineer at Logitech)

2019 - 2022

MSc in Computer Science at the Swiss Federal Institute of Technology Lausanne (EPFL) with a specialization in Machine Learning and Digital Signal Processing.
> Master's Thesis at Logitech (see above)
> Percussion Sample Classifier (semester project in the LCAV)
> JUCE VST audio plugin (optional semester project in the LCAV)
> Twitter Sentiment Analysis (CS-433 - Machine Learning)

Extra-curriculum activities:
> President for CLIC - CLIC is the association of 1500+ Computer and Communication Sciences students at EPFL. My position involved management and decision making, organisation and leadership (12 committee members), regular follow-up with members and commissions (5 commissions), external representation in particular with the main EPFL institutions, and much more.
> Sponsoring Manager for CLIC - My position involved regular contact with our sponsors and search for new partners, contract negotiations, internal collaboration with the executive committee, weekly meetings, and participation in important decisions of the association.
> Student representative for Master students - Participated in semester meetings including the IC School management and administration. Presented and discussed issues as well as potential course and curriculum improvements. Also discussed the students' official feedback and suggested solutions to the IC School. Several of them were implemented.
> Student Assistant in CS-101 ($) - Helped students solve the weekly exercises and prepare the "Advanced Information, Computation, Communication I" exam. I did that twice, in 2017 and 2020.
> Communication Staff for the EPFL IC School ($) - live social network coverage of various EPFL events (Applied Machine Learning Days, EPFL Open Days, talks, workshops, ...). This involved technical understanding of the scientific subjects, french and english skills, communication skills, and knowing how to popularise science.

2015 - 2019

BSc in Computer Science at the Swiss Federal Institute of Technology Lausanne (EPFL) with a specialization in Software Engineering.
> 3D Indoor Forensic Using a Drone (semester project in the LDM)
> Transmission over a Continuous-Time Channel with Impairements (COM-302 - Principles of Digital Communications)

Extra-curriculum activities:
> Event Manager for CLIC - My position involved the organisation of +10 events including the IC School Supper (+400 participants) and the IC Boost Day (a career day specifically designed for IC students), logistics (purchase, storage, installation), weekly meetings, and participation in important decisions of the association.
> First year student Coach - This position involved welcoming the first-year students in their new university, and organising events (aperitifs, games, etc). I mentored a group of 10-15 first-year students on a regular basis to talk about courses, advices, and student life in general.

2015

Baccalauréat S (major in Mathematics, European section program) with the highest honours, at the Lycée Français Jean-Mermoz of Dakar, Senegal.

Additional Certificates

Machine Learning Engineering for Production (MLOps) (in progress)

Introduction to Machine Learning in Production

Machine Learning Data Lifecycle in Production

Machine Learning Modeling Pipelines in Production

Deploying Machine Learning Models in Production

TensorFlow: Advanced Techniques Specialization (Sept 2023)

Custom Models, Layers, and Loss Functions with TensorFlow

Custom and Distributed Training with TensorFlow

Advanced Computer Vision with TensorFlow

Generative Deep Learning with TensorFlow

Tensorflow Developper Specialization (Dec 2022)

Introduction to TensorFlow for Artificial Intelligence, Machine Learning, and Deep Learning

Convolutional Neural Networks in TensorFlow

Natural Language Processing in TensorFlow

Sequences, Time Series and Prediction

Deep Learning Specialization (Nov 2022)

Neural Networks and Deep Learning

Improving Deep Neural Networks: Hyperparameter Tuning, Regularization and Optimization

Structuring Machine Learning Projects

Convolutional Neural Networks

Sequence Models

Personal Projects

Spoteezer I love music, and sharing music across platforms like Deezer and Spotify has always been a struggle. That's why I created Spoteezer - a web app - converting Spotify links to Deezer (v1) and back (v2). Built with HTML/CSS (frontend) and a Flask Python server (backend), it interacts with both platforms' APIs to return the best match.
> Try it out!
> Code

Demo

French Intent Classification Intent classification is a process used in NLP, e.g chatbots, where the goal is to determine the intent behind a user's text input. It involves categorizing a piece of text, like a sentence, into a predefined category or intent. As part of a technical challenge, built a French Intent Classification application using pre-trained HuggingFace Transformers models (FlauBERT, CamemBERT) as word/sentence embeddings, and classic ML algorithms (Logistic Regression, MLP) as classifiers. To train the classifiers, I used the CLINC150 dataset. The training and validation sets were sampled to only keep the 9 classes of interest, and translated to french before training the classifiers. I obtained an average F1-Score of 98% on the test set with the best model (CamemBERT Large + Logistic Regression). This is - for now - the open source repository reflecting the most my current project structuring skills and coding style.
> Code

Demo

Forum EPFL Search While looking for companies to contact during my job search, I realised that the Forum EPFL website's search engine was not very practical as it only compared the search query to company names. Therefore, I started a project doing the following:
1/ Scrap the Forum EPFL website to gather all the companies' information, job offers, and seeked profiles;
2/ Generate a prompt based on the user's profile and interests;
3/ Find the best matching companies by giving the user-tailored prompt and the scraped data to a Large Language Model;
> Code (WIP)

Master's Graduation Video I extensively participated in the ideation and the editing of a video project to celebrate the 2022 graduation of EPFL Computer and Communication Sciences Master students. This 15-minute video was broadcasted during the ceremony gathering the students and their guests.
> Watch it here!

This website The idea is to gather all my work and projects in one place. I built this website using HTML and CSS only based on Andrej Karpathy's website as template, whom I thank for the inspiration! This was a great opportunity to learn about web development and some design, as well as CI/CD pipelines using github actions. Feel free to reuse the template! (warning: the CSS file is a mess)
> Code

My Music

I started producing electronic music in 2018, first with Cubase, then FL Studio, and finally Ableton Live. This is a long process of listening to others' work, educating my ears, and learning/practising the tools in order to compose tracks that I like and sound professional. I compose, sound-design, mix, and master everything myself. Here are some of the tracks I released and I am proud of:

D/WN · 4:44

D/WN · phi

D/WN · SPLIT

D/WN · BOUNCE

Follow me on your favorite platform!