Ayush Kumar Shah
Machine Learning Engineer,
Computer Science Instructor,
Hello! It’s me, Ayush Kumar Shah, an Artificial Intelligence enthisiast. Currently, I am active in working with global client teams to build state-of-the-art products. I have worked in the domains of Recommendation System, Handwritten character recognition system, waste classification system, Computer Vision Systems during my time at Fusemachines.
My inquisitive nature, craving for knowledge, and longing for novelty and innovation strengthen my passion to work and learn daily to increase my knowledge horizon. I am mostly into python and AI, so my blog will be a reflection of whatever new thing I learn about them. Thank you for visiting my blog.
Check out my short resume below. You can view or download the complete pdf version of my CV here: CV_AyushKumarShah.pdf
BE, Computer Engineering
CGPA: 3.96 / 4.0
Coursework: Machine Learning; Artificial Intelligence; Natural Language Processing; Data Structure and Algorithms; Database Management Systems; Operating Systems; Advanced Calculus and Linear Algebra; C; C++; Computer Architecture; Software Engineering; Digital Signal Processing; Embedded Systems
June 17, 2019 - Present
Machine Learning Engineer
Aarya Tower Bhawan, Hattisar, Nepal
Working on several client-based ML projects from the US as well as internal ML projects of the company in the field of NLP, Computer vision, Recommendation systems. Worked on the whole ML pipeline: data collection, data cleaning and pre-processing, model building, tuning hyperparameters, model training, and model evaluation. Some of the projects I have worked in are:
Nepali Intelligent Character Recognition System
A framework built by combining RNN and CNN to predict the handwritten or printed texts in both English and Nepali language in different fields of a form. Worked on building binary language classifiers and a complete prediction pipeline for texts in images of different types of forms.
Recommendation System for a subscription-based e-commerce client to increase revenue through cross-selling. Involved in data wrangling, feature extraction, model experimentation, and implementation. Techniques used: Item-to-Item models, (Neural) Collaborative Filtering, Factorization machines, etc. Increased revenue by 6% (large as we were serving 600k users).
Automatic classification of products on the basis of various chemical attributes to optimize business decisions for products that go unsold. Worked on Boosting algorithms like Gradient Boosting, Random Forests, XGBoost, LightGBM and so on for regression and classification tasks.
Involved in design, review, and refinement of content - reading material, quizzes, assignments, and projects for Fusemachines AI Education Programs - “Foundations in AI: Mathematics for AI”, “Micro Degree™ in Artificial Intelligence: Machine Learning, Computer Vision”
Jan 2020 - Present
Computer Science Instructor
- Teaching the course “Foundations in AI: Computer Science and Mathematics” offered by Fusemachines to undergraduate BSc.CSIT students.
Nepali Plagiarism Detection (NLP)
An application that detects plagiarised Devanagari text files using a self-built rule-based stemming algorithm and Cosine similarity on TF-IDF vector representations of the documents.
Guitar chord recognition app
An application that predicts the chords when the Mel spectrograms of guitar sound are fed into a CNN.
Sarangi: Nepali lyrics emotions extraction (NLP)
A framework that categorizes songs written in the Devanagari script into four emotions using Naive Bayes and SVM.
AI Plays GTA 5: Simulating self-driving vehicles
A bike-riding agent in a virtual environment (GTA5), built using CNN, used for simulating self-driving vehicles.
MathMate – advanced mathematical calculator
An android app that solves different types of mathematical equations, numerical computations, and calculus problems showing involved steps
ADDITIONAL EXPERIENCE AND CERTIFICATES:
- Computer Vision Nanodegree - Udacity
- Deep Learning Nanodegree - Udacity
- Natural Language Processing Nanodegree - Udacity
- Python: Design Patterns - Linkedin Learning
- Full Stack Web Development with Flask - Linkedin Learning
- Pandas foundations - Datacamp
- Software Engineering for Data Scientists in Python - Datacamp
- Multivariable calculus - Khan Academy
- Statistics and Probability - Khan Academy
- Fusemachines Artificial Intelligence Scholarship Program (Nov 2018 - May 2019)
LANGUAGES AND TECHNOLOGIES:
- Proficient: Python, Numpy, Pandas, Matplotlib, OpenCV, nltk, Keras, TensorFlow, Pytorch, git
- Familiar With: C, C++, SQL, AWS, LaTeX