Skip to content
View pptr3's full-sized avatar
:octocat:
Doing stuff for achieving other stuff
:octocat:
Doing stuff for achieving other stuff

Highlights

  • Pro

Block or report pptr3

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Please don't include any personal information such as legal names or email addresses. Maximum 100 characters, markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
pptr3/README.md

👋 Hi there!


class Petru:
  def __init__(self):
    self.name = 'Petru'
    self.surname = 'Potrimba'
    self.education = 'Master Degree in Artificial Intelligence and Bachelor Degree in Compuer Science and Engineering'
    self.experience = ['Roboflow', 'AIvolution', 'Sis.ter', 'Imola Informatica', 'Unibo']
    self.roles = ['Machine Learning Engineer', 'Researcher', 'Software Engineer']
    self.interests = ['Machine Learning', 'Deep Learning', 'Competitive Programming', 'Teaching', 'Communication']
    self.languages = ['English', 'Italian', 'Romanian', 'Spanish']
    self.hobbies = ['Tv series','Swimming', 'Fitness']

🔨 Most interesting projects


  • Personal Blog : Actively write blog posts about my research projects mainly focused on Software Engineering and Machine Learning.

  • Send Time Optmization : Send Time Optimization (STO) is an add-on feature that allows marketers the ability to send emails at the most optimal send time to each recipient. We used feature engineering to create meaningful features and then we used deep learning to address this problem. The implemented solution improved the ratio of opened emails by ∼ 32% and the click ratio by ∼ 12%.

  • Generate synthetic faces : Auxiliary classifier generative adversarial network (AC-GAN) model that can generate synthetic conditioned faces with a specific set of attributes (example: generate a face of a woman, with blond hair that smiles). The FID metric reached a score of 19.42 for the generation of conditioned synthetic faces and a score of 11.07 for the generation of random synthetic faces.

  • Math assistant : Implemented and deployed a custom Optical Character Recognition (OCR) within an iOS Mobile application and a Telegram bot for the recognition of handwritten mathematical expressions. Implemented a Machine Learning model that understands whether the mathematical expressions are correct or not.

  • Image captioning with Transformers : Implemented the Image Captioning model using deep learning with Transformers. The model has been trained both on COCO and Flickr8k datasets. The performance reached results comparable to the state-of-the-art of 2020 on the BLEU and METEOR metrics.



🎓 Academic achievements

  • Top 0.4 % student among all Masters students of the Engineering department of Bologna University.
  • Winner of Start Hub competitive programming coding challenge (first position among 200+ competitors) link.
  • Admitted to ETH University for Computer Science Master’s Degree (ETH engineering department is 4° ranked among all the engineering departments all over the world).

📫 Connect with me


LinkedIn Gmail Blog GitHub

Pinned Loading

  1. math-assistant math-assistant Public

    Mobile (iOS) application for analysis and recognition of handwritten mathematical operations. The goal is to understand if the operation is correct or not. If it is not, it shows you the correct re…

    Swift 3 1

  2. machine-learning-ng machine-learning-ng Public

    My personal solutions in Octave/MATLAB and Python at the assignments of Machine Learning course on Coursera held by Mr. Andrew Ng. -> https://www.coursera.org/learn/machine-learning

    Jupyter Notebook

  3. connecting-rods connecting-rods Public

    Computer vision system aimed at visual inspection of motorcycle connecting rods.

    Jupyter Notebook

  4. road-sings-recognition-pytorch road-sings-recognition-pytorch Public

    Multiclass classification model for the recognition of 21 different road signs used in autonomous car systems using pytorch.

    Jupyter Notebook

  5. face-generation-gan-keras face-generation-gan-keras Public

    Auxiliary classifier generative adversarial network (AC-GAN) model that can generate synthetic faces conditioned with a specific set of attributes. The FID metric reached a score of 19.42 for the g…

    Jupyter Notebook 1

  6. image-captioning-keras image-captioning-keras Public

    Image captioning using Bahdanau and Luong attention applied on COCO and Flickr8k datasets. The performances have been measured using BLEU and METEOR scores.

    Jupyter Notebook 2