Riccardo El Hassanin

I'm

About

Electrical & Electronic Engineer

  • Nationality: Italian
  • City: London, United Kingdom
  • Email: riccardo.elhassanin27@gmail.com
  • Phone: +44 (0) 774 7712 087

I am a multi-disciplinary engineer with a strong foundation in machine learning, statistical modelling, and software development. I'm passionate about building and optimising quantitative models and developing high-performance, scalable solutions by combining technical expertise, analytical thinking and research skills with creative problem-solving. I'm driven by continuous learning and innovation in novel machine learning and deep learning techniques across fields such as healthcare and finance. I'm very inspired by the transformative potential of automation and AI in shaping the future. I thrive on tackling new challenges and I always look forward to new endeavours. In my free time I enjoy working out at the gym and playing basketball.

Education

MEng in Electrical & Electronic Engineering

2019 - 2023

Imperial College London

I have received my MEng degree in Electrical and Electronic Engineering from Imperial College London, where I achieved First-Class Honours (70+/100). I focused my studies on machine learning, deep learning, artificial intelligence, embedded systems, algorithms and data structure, computer vision, pattern recognition, large data and signals processing, and computational finance.

International Baccalaureate (IB) Diploma

2015 - 2019

GEMS World Academy Dubai

Total Score: 43/45 | Higher Level Subjects: Mathematics (7/7), Physics (7/7), Business Management (7/7).
Salutatorian Award (2019) and Honours Roll Awards (2015-2019) for outstanding academic performances.

Work Experience

Data Scientist | ML Engineer - Internship April 2024 - July 2024

Agreena

Developed and optimized deep learning models for large-scale data analysis, focusing on satellite scene quality classification (using optical satellite imagery from Sentinel 2) and deployed them to production environments with VertexAI and Google Cloud services. Conducted research on cutting-edge technologies (CNN, EfficientNet, ResNet, HuggingFace Vision Transformers) to optimize and improve model performance by 6% reaching 97%. Obtained hands-on experience in extracting, cleaning, and preparing large-scale remote sensing data for ML products, enhancing data quality, integration, and functionality across diverse inputs and models. Wrote production grade code and formulated unit, integration and end-to-end testing frameworks.

Machine Learning Engineer - Contractor Oct. 2023 - Jan. 2024

FireX.ai

Engineered deep learning models using multi-modal remote sensing datasets (optical and SAR satellite imagery) for predictive modelling to aid the assessment of wildfire risks for environmental risk analysis and disaster mitigation. Gained practical experience in implementing complex algorithms and leveraging deep learning frameworks in cloud environments, writing high quality and modular production grade code (with unit, integration and end-to-end testing) in a fast growing start up.

ARM® Project Consultant May 2022 - July 2022

Imperial College London - ARM

Led a team of 6 to build an open-source software and data processing model to perform local speaker recognition in real-time, which correctly identified 90% of speakers in roughly 1.3 seconds per inference. The project was provided by ARM® to develop a speaker recognition platform that could be incorporated into larger applications. Developed a low-latency Python-based algorithm to extract acoustic features from voice spectrograms using a CNN-based model and a cosine-similarity metric to match the speaker to a person in the database, significantly enhancing speaker identification through robust data analysis, image processing, and predictive modelling techniques.

Undergraduate Researcher June 2021 - Sept. 2021

Imperial College London

Undertook a research experience, supervised by Professor Tom Pike. Applied advanced statistical analysis and data processing techniques to processed data from NASA’s Mars' InSight mission to disentangle the seismic signals from Martian environmental interferences and uncover meaningful patterns. Created a UTC-Mars time-converter for the purpose of mapping the processed data. Contributed to team research and enhanced communication skills delivering reports, collecting results and research findings.

Tutoring IB students: HL Math and HL Physics Aug. 2020 - Dec. 2020

Guided penultimate and last-year IB students through practice problems for HL Maths and HL Physics classes, answering any questions about concepts and next steps. Created lesson materials to improve comprehension of HL Maths and HL Physics topics such as practice worksheets.

Projects

LLM RAG Chatbot

Developed a Retrieval-Augmented Generation (RAG) chatbot capable of answering questions sourced from provided documents, leveraging LLaMa 3.1 open-source Large Language Model (LLMs) for real-time, dynamic conversational AI. Utilized LangChain and vector databases to efficiently perform document embedding and similarity search, optimizing the retrieval of domain-specific information and enabling accurate response generation based on relevant knowledge bases. Implemented the system using Flask and FastAPI to run the chatbot on a web browser, enabling user-friendly interaction and efficient deployment.

MEng Final Year Project 2023

My master thesis at Imperial College proposes a Complex-valued Deep Canonical Correlation Analysis (CDCCA) model to study the intrinsic relationships between the components of two financial time series data. The model aims to learn complex nonlinear transformations of two datasets such that their canonical correlation is maximised. It can be viewed as an extension of the parametric Deep Canonical Correlation Analysis (DCCA) that leverages the expressive power of complex numbers in order to deliver deeper insights. This complex-valued model showed superior performance, highlighting its potential for multi-view learning tasks.

Speaker Recognition Project

Created an open-source software and data processing model to perform local speaker recognition in near real-time. Trained an advanced convolutional neural network to extract the acoustic feature information from spectrograms of people voices. Developed an algorithm in Python that runs on a Raspberry Pi4 to extract the input voice features using the trained CNN and match the speaker’s input voice to a person in the database. Correctly identified most speakers in roughly 1.3 seconds per inference.

Music Synthesiser

Wrote an embedded software in C++ to make a music synthesiser work. The system was implemented using interrupts and threads to achieve concurrent execution of tasks, such as detecting key presses and generating the output waveform. Used a mutex and queue for thread-safe data sharing. Implemented the ability to send and receive messages to other keyboard modules (when stacked together) using the CAN bus. Advanced features: waveform selection to change the timbre, polyphony & handshaking.

Sound Recognition Device

Notisound is smart sound recognition device that classifies doorbell and fire alarm noises in a house enviroment and sends notification to the end user through an app when the sounds are detected. This embedded device uses: a signal-processing algorithm to pre-process the microphone inputs into spectrograms, a convolutional neural network to constantly classify the sounds and a mobile phone app to receive the notifications.

Mars Rover Project

Worked in a team of 6 to design and build an autonomous rover system that could be used in a remote location without direct supervision. The rover is capable to receive movement commands and send status data. The rover sucessfully detects and avoids obstacles in its working area and it is able to build a map and store data of its travelled path. Hardware used: Arduino, DC motors, FPGA, ESP32, camera, SMPS.

CPU Design Project

Worked in a team of 4 to design a general-purpose CPU using Quartus Prime and Verilog HDL, able to execute three algorithms: calculating Fibonacci numbers, calculating pseudo-random integers using a linear congruential generator and traversing a linked list to find an item. The CPU follows a Harvard architecture design in order to implement the desired algorithms.

Skills

Technical Skills

Microsoft Office Applications
Python (Pytorch, TensorFlow, NumPy, Pandas, Scikit-learn)
C++
Matlab & Simulink
Linux
WEB DEVELOPMENT: HTML, CSS, Javascript, Bootstrap
HARDWARE SKILLS: Raspberry Pi, Arduino, FPGA
Intel Quartus Prime & Verilog HDL
Cloud Computing: AWS, Google Cloud
LTSpice: circuit simulator
Autodesk Fusion 360 / AutoCAD
Soldering, 3D Printing & Laser Cutting

Language Skills

Italian Native

English Full professional

Spanish Intermediate level

Contact

Should you require further information please do not hesitate to contact me.

Location:

London, United Kingdom

Email:

riccardo.elhassanin27@gmail.com

Call:

+44 (0) 774 7712 087