chubakbidpaa@gmail.com or
nazemi@atazir.net.au*
Chubak#7400
https://github.com/chubek |
https://chubek.github.com
https://www.linkedin.com/in/chubak-bidpaa-0a3798189/
PDF Version
Machine Learning Engineering, Artificial Intelligence, Automation
Freelance projects, sorted in order of importance:
| Project | Skill | Experiences | Language | Tools and Techs |
|---|---|---|---|---|
| Mo Face Recognition and Liveness Detection | Vision | Using Keras and DeepFace to create a liveness detection and facial recognition network. Using Flask to create a fluid backend for a face recognition project. Assessing image quality Did successful unit tests across all modules. Good API practices such as logging and return codes. Creating facial embedding vectors. Using jQuery to send requests for a simple frontend. Validating user settings for errors. Successfully used a custom network for liveness detection with an accuracy of 97%. Wrote adequately detailed documents. |
Python, HTML, JavaScript | Keras DeepFace MTCNN Dotenv Flask imgaug OpenCV Image-Qualiy Flask Numpy |
| SparrowVoice Convertor | ML Engineering | Using GMM models to convert between voices. Using intermediate GUI systems to create a GUI in Python. Using PyDub, SciPy and Numpy to process sound samples. Extracting feature statistics from sound files. |
Python | Sprocket-VC, PyDub DearPyGUI, PyWorld Scikit-Learn SPTK SciPy Numpy |
| Frontend for OpenAI Jukebox | ML Engineering | Using Jupyter notebook forms to create a GUI. Creating a sound environment for the end user. Dissecting other people's projects. |
Python | Jukebox Jupyter Notebook forms |
| London Tram Arrival Time Predictor | Shallow Learning | Using XGBoost to create a regression model. Selecting the best hyperparameters using Hyperopt paramter optimization. Preparing the script for the end user. |
Python | XGBoost Hyperopt Scikit-Learn Pandas |
| Various Automation tools | Automation, Shallow Learning | Using Selenium and Playwright to automate the browser. "Hacking" the endpoints to receive the target data. Creating CLI tools for the end user. Using OCR to read characters from screenshots. Using Google Maps to get GeoJSON data. Using Docker to containterize and serve systems. Using K-Means clustering to cluster information. Using lxml and XPath to get HTML elements Using Google Cloud Platform APIs to serve the means of the project. Storing data in BigQuery and SQL tables. Doing SEO with Google Analytics and backlinks API. Using Regex patterns to validate data. Used RAKE to get keywords from text. |
Python, Go | Selenium Playwright HTTP tools Base64 tools OCR tools XPath lxml Google Cloud API tools Scikit-Learn Regex Tkinter Rake Docker Joblib Pandas PyInstaller Go and Python MySQL and MSSQL tools Azure Google Cloud Platform Go Excel tools Google Analytics |
| Reddit Scraping and Automation Tool | Web Scraping, Automation | Using PRAW to scrape and automate Reddit. Using Selenium and lxml to scrape pages. |
Python, JavaScript | PRAW Selenium lxml Docker Google Cloud APIs (BigQuery, Programmable Search, Sheets) |
| Potpourri SEO Analyzer | Web Scraping, Automation | Successfully creating a Python package from scratch. Applying object-orieted programming practices. Search Engine Optimization analyzer. |
Python | Google API lxml requests regex Concurrency and Parallelism NLTK RAKE BS Parser |
| Forex Trader Bot | Deep Learning, Automation, Deep Reinforcement Learning | Trained a classifier that predicted short/long. Trained a regressor that predicted the change in Close indicator. Trained a reinforcement learning model that traded based on technical analysis indicators and portfolio metrics. |
Python | Python's ta library (Technical Analysis) keras-rl2 Tensorflow and Keras Google Colab Sklearn |
| GPT-3 Wrapper | Fullstack and ML Eng | GPT-3 Wrapper for Generating Articles | Python, HTML, JavaScript | FastAPI, OpenAI GPT-3 Hugging Face Jupyter Notebook Google Colab TheFuzz (Formerly Known as FuzzyWuzzy) |
| Schoolingo Document Scanner | Computer Vision, OCR | Used classical computer vision to scan documents. Used Tesseract to do OCR. Used Spacy's NLP tools to filter certain parts of the document. |
Python | Spacy OpenCV Tesseract |
| Various Deep Learning Training Tools | Deep Learning | Using FiftyOne toolset to sample data. Using Unet to segment imags. Training an original network for Liveness detection. Using Google Colab Pro for cloud computing. Using Google Drive in a novel way. Doing NLP with GloVE embedding vectors. Using original Conv2D and LSTM networks to regress and classify data. |
Python | FiftyOne Keras UNet Torch Scikit-Learn OpenCV PIL Numpy SciPy BERT Keras-Segmentation GloVE |
| Adobe Premiere MOGRT Automation Tool | Automation | Successfully used ExtendScript and CEP panels to get MOGRTs from a Premiere project and set it
back. Made a concurrent backend with Sandboxie. |
JavaScript, Node.js, ExtendScript, Python | Sandboxie Flask JSON File IO |
| A series of Machine learning exercises for a course | Shallow/Deep Learning | Writing a Gradient descent algorithm from scratch. Creating an basic Autonomous driving system from scratch. Writing over 60 ML/DL exercises. Using matplotlib to visualize data. Writing loss functions, cost functions, and objective functions from scratch. |
Python | Self-created tools Keras Scikit-Learn Pandas Wget Jupyter Notebook Linear Regression Ensemble Methods Logistic Regression Clustering methods Decision trees Support Vector Machines. |
| Web Scraping with Rust | Web Scraping, Automation | Successfully used Rust for browser automation and web scraping. | Rust | Reqwest Tokio Fantoccini Hyper Serde Regex |
| A CRUD server in Rust | CRUD | Successfully created an asynchrnous web server with Rust that stores data in MongoDB.M Used RAKE to get keywords from texts. |
Rust | Rake Tide async-std Serde mongodb |
| PDF Table Comparator | Automation | Successfully extracted tables from PDF with titles. Matched tables based on titles and compared them. Generated dynamic HTML of this comparison results. |
Python | camelot Tesseract Bootstrap FuzzyWuzy Pandas |
Personal Projects, sorted in order of importance:
| Project | Skill | Experiences | Language | Tools and Techs |
|---|---|---|---|---|
| markdown2docx | Automation | Successfully created Markdown parser in Golang Used Regex to match Markdown syntax |
Go | Regex Unioffice |
| Will Sh3 B33 | Shallow Learning, Deep Learning, NLP | Trained 2 shallow models on OkCupid data. Traind a deep model on OkCupid data. Fine-tuned BERT on OkCupid essays. Created a backend, a frontend, and deployed the model on a Docker container. |
Python | Tensrofflow and Keras Scikit-Learn Docker Ubuntu BERT and BERT Tokenizer |
| Bedlam Noise Apparatus | Adobe After Effects Plugin | Successfully used Adobe After Effects API, OpenGL and GLSL to create a plugin that creates six types of noise. | C++, GLSL | OpenGL After Effects API |
| Mongoose Jumblator | Encryption | Successfully made a Mongoose plugin which encrypts MongoDB entries and decrypts them back. Did successful unit tests. |
TypeScript | Lodash Mongoose Crypto-JS Chai Mocha |
| The ML Codex | Education | Wrote a few chapters of a book using LaTeX about implementing custom Machine Learning algorithms. | LaTeX | Diction English Fluid Language Helpfulnesss LaTeX skills |
| Syren Digital Signage | CRUD | Successfully created a digital signage backend with Kotlin. | Kotlin | Spring Framework |
| Mathcord | Automation, Cryptography | A Discord bot that uses Shunting Yard to calculate mathematical expressions, with SHA-512 and ED25519 built from scratch | Go | Only Go Standard Library |
Note: Most of these projects, personal or commercial, if open-source, can be found in my Github. Find the link in the resume header.
Inflation of the host country is applicable in fees.
Please inform me through other means if you use the Atazir email.
Last Updated: 10/02/2022 (dd/mm/yyyy)