Charles Vin

Freelance ML Engineer

Me
github_icon bsky_icon linkedin_icon mail_icon

Project Spotlight

Paper re-implementation: ClimODE: Climate Forecasting With Physics-informed Neural ODEs, ICLR 2024
PyTorch Torch Ignite CNN ViT ODE

Neural Network DIY

  • Development of a neural network library entirely in Numpy
  • Implementation of essential modules (linear layers, 1D convolutions, etc.) for creating, training and evaluating neural networks.
  • Performance optimization through advanced use of Numpy for efficient computation
Numpy Mathematics Neural Networks

Speaker Recognition & Sentiment analysis

  • Design and implementation of a natural language processing system based on bag-of-words representations for speaker recognition and sentiment analysis.
  • Use of textual data pre-processing and machine learning modeling techniques (Naive Bayes, logistic regression, SVMs).
  • Management of class balancing and estimation of model generalization performance.
NLP Bag of word Unbalanced learning
See more

Experiences

Filter by roles:
Filtered

  1. ML Engineer - National Audiovisual Institute (INA)

    • Design and implement an image search engine unlocking access to 250K+ hours of archival video content, surfacing previously hard-to-discover media.
    • Orchestrate large-scale on-premise embedding pipelines with Airflow DAGs and vLLM, producing more than 1 billion image embeddings.
    • Architect and operate large-scale data pipelines (ETL) processing terabytes of multimodal content, including semantic transcription chunking, embeddings, and metadata, enabling high-performance, low-latency Elasticsearch indexing at scale.
    • Deploy and optimize vLLM inference infrastructure on k3s/k8s using Helm, benchmarking performance against local pod-based inference to maximize scalability and efficiency.
    • Conduct in-depth research on vector quantization techniques in vector databases (e.g., Qdrant, Weaviate), benchmarking their performance in Elasticsearch and OpenSearch to inform system design decisions.
    • Drive full-stack development across frontend, backend, and search infrastructure using Vue.js, FastAPI, and Elasticsearch.

  2. Research Internship - Self-supervised Learning on Satellite Image Time Series

    • Conducted extensive literature review on state-of-the-art methods in self-supervised learning (SSL) applied to Satellite Image Time Series (SITS).
    • Reproduced and benchmarked baseline models to ensure robust performance comparisons.
    • Developed and implemented an innovative image retrieval pretext task utilizing a specialized hierarchical loss function to train SSL models on SITS data.
    • Actively participated in research lab activities.

  3. Research project - Explainable insects classification

    • Development of a convolutional neural network for insect classification.
    • Use XAI with gradient-based approaches and the LIME and SHAP frameworks.
    • Takeover, adaptation and documentation of the project.
    • Experiences monitoring and evaluation of different models with WandB.

  4. Research Project - INRIA

    • Study of the trade-off between observation and action in reinforcement learning (subject details)
    • Understanded, independently, a cutting-edge problem in reinforcement learning.
    • Demonstrated perseverance and determination in the face of technical challenges.
    • Implemented an experimental plan.

Education

  1. Master's Degree in Computer Science - Specialisation Data, Machine Learning, and Knowledge (DAC)

    Relevant courses:

    • Image and signal processing
    • Mathematics for ML (statistics/probability, Markov Chain, ...)
    • Databases: SQL, XML, JSON, distributed databases
    • Machine Learning: classification, neural networks, decision trees
    • Natural language processing, information retrieval
    • Opening courses from Master of Mathematics: statistical learning and convex optimization.
    Learn more

  2. Bachelor's degree, Mathematics and Computer Science applied to Cognitive Science

    Grade: Highest Honours

    Relevant courses:

    • Decision statistics: practical work in R.
    • Probability and measurement theory
    • Cognitive science

    Relevant academic projects:

    • Bayesien Tweet Classifier: "Positive" or "Negative" tweets classifier
    • Basic ML algorithm from scratch: KNN, KMean, linear/polynomial regression
    Learn more

My main sports

Running

Cycling

For commute and recreation

Climbing

Last 4 weeks

24 km

Distance

135 min

Moving Time

Total Activities

Fetched using Strava API

Hobbies

Me & Cooking

Let me introduce my cooking philosophy throught a few key points

  • Passionate Home Cooking: I cook often at home, driven by a deep passion for creating incredible food.
  • Relaxed Approach: I don't sweat the small stuff. I'm all about enjoying the process and flavors, not stressing over perfection.
  • Feeding Loved Ones: I love cooking for my family and friends, and I believe there's nothing quite like a homemade meal to show you care.
  • Budget-Friendly Goodness: I'll show you how to make delicious meals that won't break the bank.

Inspiration

I drew my inspiration and passion from a multitude of talented YouTubers. Here's a curated list of some of the best who have played a significant role in shaping my culinary journey:

Plants & Gardening

I've always had a deep appreciation for nature, growing up in the countryside helps a lot. However, my passion for plants truly blossomed during a high school biology class when I witnessed chloroplasts moving under a microscope. Those tiny green organelles, brimming with life and releasing oxygen bubbles, captivated me.

My journey into the world of plants began in my high school years, largely influenced by YouTube. Over time, my interests expanded to gardening, agriculture, and eventually led me to explore "self-sufficiency" YouTubers. These individuals are dedicated to achieving energy and food self-sufficiency (without delving into extreme survivalism). It all intertwined with my fascination for culinary pursuits like making jams and preserving food.

My passion for music, particularly electronic music, ignited during my middle school years when I stumbled upon YouTube channels like MrSuicideSheep and Monstercat. From that moment on, I never stopped listening. I take pride in exploring the entire spectrum of electronic music.

If you're curious about what's currently playing in my world, I invite you to check out my Last.fm profile, which aggregates my Soundcloud listens and gives you a glimpse into my musical journey.

My Last.fm