My Portfolio

My name is Chelsea, I study concepts and technologies in the field of AI and ML with the vision of one day having my own company that specializes in Virtual Reality & Computer Vision. On the process of getting there, a lot of skills has to be learned, either for the purpose of attainment of that definite goal or as a way to offer job recruiters my services.

As easy an answer as possible, I thought that the best way to show what I can do is to actually learn those skills and practice on them in terms of short or long term projects. Short term projects are perhaps those that you want to add up in your portfolio while long term might be the ones that complement your overall plans for the(your) future.

Below are a few of the skills I've managed to gather within a few months. What I want the most? To be sufficiently equipped with enough skills, networks, ideas and money that would push me steps closer to my big goal.

Skills

• HTML
• CSS
• Web-Development
• Machine-Learning Algorithms

• Tensorflow
• Javascript
• Git
• Google Cloud Platform
• Python

• SEO
• Virtual Reality
• Writing & Research
• React

Certifications

1. I consider myself a near expert in Machine Learning(Algorithms). I can use and implement all popular ML Algorithms to build systems that can make predictions at a minimal level. Ive practised extensively with ML Algorithms mainly during my Machine Learning Online course on Coursera I completed the course with a score of 95%. Below is an image showing my Certificate and a link that you can verify the results directly on Coursera.

Certificate link:Machine Learning - Standford University



2. I can use Tensorflow for classification or regression problems. I understand the loss function, optimizer, model fit and all the important functions that are necessary for building a model. Using Keras as the back-end library i can do data preprocessing with numpy, callbacks, evaluation and a few other things. Ive had practice using Tensorflow on the MNIST dataset for HandWritten Digits classification and MNIST dataset for Clothing items classification. I have a Certificate on Tensorflow for Basic Image Classification, a Guided Coursera Project. Image and direct link to Coursera below.

Certificate link:Basic Image Classification with Tensorflow



3. Still on Tensorflow, but way more advanced now. I love to use Tensorflow bescause of it's ease-of-use and extensive applications. I'm currently taking a Tensorflow specialization on Coursersa Learning platform Tensorflow In Practice. It's a great course, that basically introduced me to Tensorflow and related libraries. As I go along in the course, we(students) complete several practical projects that deepens our understanding of the concepts. I am more concerned with real-world applications of the skills I learn. I'm constantly thinking of ways to apply them. With Tensorflow, I know I can build seamless ML & DL applications and look forward to doing so. Some of the skills I've learnt and currently applying in a few projects are:


• Regularization with Dropouts, Batch Normalization
• Data Augmentation to Reduce Model Overfitting
• Image Generator Library to Automatically label classes of images from a dataset
• Techniques for Processing Images before Training
• Data Cleaning with Python and Plotting with Matplotlib
• Serving Tensorflow Models for production with TFX API and Docker Images
The listed skills have been learnt during online courses and while implementing personal projects. Below is the link to Certificates from completion of Course 1 in Tensorflow in practice specialization

Certificate link:Tensorflow for AI, ML & DL

Projects

I have applied various skills gathered to the implementation of different projects, specifically in the filed of Deep learning. My interest lies in Computer Vision and Virtual Reality fields. The main reason why I build projects is becasue I love to discover real-world applications of any skill I am learning. that was what drove me to build two different websites back the when I was learning web-development. At this point in time, I consider myself solely a beginner in CV & VR but one that definitely makes sure to learn and move a step further each day. I build projects for two main reasons: To develop a good understanding of concepts as I implement them, and secondly, to scale meaningful projects into production services. Moving on, these projects listed inherently describes what I can and have done as a developer, and of course as I learn and develop more expertise, I hope to build applications that would be invaluable to customers and people in general.

• A Simple Web App to Classify Images

  One of the earliest projects I worked on excluding the assignment from online courses. In this project I used most concept I had learnt to build a simple web app that can predict the classes of images. I basically built the model with Tensorflow and keras, on google Colab and served using Tensorflow Serving built on Docker. Some Technical Information on this project: In this project, I used the Cifar100 dataset from 2014 developed by to build an Image classifier. This dataset consistes of 50000 images distributed into 100 different classes. I faced some overfitting when implementing this, my accuracy was low at first, this made me learn some new techniques that helped resolve that and consequently bumped up accuracy. The technique applied were generally Regularization and Data Augmentation. For full detailed code, refer to my github repo here and also a blog post where I taught how to achieve high accuracy on this dataset here

The web app was built using HTMl as frontend and Python Flask as back-end. The user here has to upload an image and the model returns the top 5 classes the image belongs to. Link to Project Implementation of Image Classification in a Web App