Personal Python projects developed in VS Code.

I have a knack for programming (it's about algorithms development, not memorization)!

Note this page is not about impeccable or professional software development — it's about curiosity, experimentation, learning and letting your creativity run free. 🐍

Python is an interpreted language that's extremely powerful and has a lot of modules that provide ready-to-use functions. It doesn't need to be memorized, except for basic syntax. It bugs me that recruiters think that a test of Python (that is mostly about memorization) is necessary. For that you can just search in the myriad of functions and libraries. Now algorithm development is different, it requires math skills and reasoning, which I have plenty of.

Various projects that I've developed in Python, including:

• AI video/image enhancement with open source code Real-ESRGAN (this is ML/DL*).
• Checking if NVIDIA's CUDA is enabled after installation.
• Speech recognition and transcription/translation (PT to EN) (this is ML/DL).
• Automatic creation of subtitle file (.srt) from transcribed/translated audio. Ana Paula
• Upward shifting of speech pitch (baby voice).
• Given a person's name in a given language, predict the gender of the person (this is ML/DL).
• Play counts syncing between iTunes and Windows Media Player.
• Moving MP3 files under their right folder and creating complete log files with changes.
• Search, download and attachment of MP3 artwork (Apple or Discogs), with log files creation.
• Image pattern recognition (trying to check if an MP3 cover is a generic vinyl/LP so it can be automatically replaced). This is done by comparing the cover with a given vinyl image (this is ML/DL).
• Searching and populating MP3 tags on the Discogs website automatically.
• Making MP3 tag names in sync with the actual file names (to detect and resolve conflicts)
• Downloading of videos from Youtube (use "YT-DLP CMD LINE.txt" to see command line usage, as Youtube credentials may be needed and are taken from browser cookies). Python codes for video download will only work if the video doesn't requires credentials. The command line works always (install module yt-dlp).
• Creation of stand-alone/static executables from Python codes.
• Installed, configured and debugged PySpark on Windows, with Python 3.11.8, Java 11.0 and Hadoop 3.3.5. This allowed me to achieve the below task (free Azure account is constrained).
• Fitting models such as Logistics Regression using PySpark ML (results below).
• Fitting ML time series models to stocks, though it's not much use if the market is not stable (in consideration).
• Web scraping of tables on Wikipedia using BeautifulSoup.
• Automatic creation of formatted PPT slides with data scraped from Wikipedia. Brazilian actresses**
• Calling function by passing another function as parameter and feeding variables to it from called function with lambda.
• Image, audio and video processing using ffmpeg.
• Submitting Windows commands from Python.

*ML: Machine Learning.
*DL: Deep Learning.

**Video created from the PPT deck created in Python (soundtrack added with ffmpeg).

Fitting of a logistics regression with PySpark to a credit risk file available on the internet.

It's a credit risk dataset (sometimes called the "Give Me Some Credit" dataset from Kaggle), used for predicting if someone will default on loans.
The binary response variable is called SeriousDlqin2yrs and indicates whether a given customer became seriously delinquent in a 2-year time horizon (1=Yes, 0=No).

The area under the ROC curve (a measure of the goodness-of-fit of the model) is modest at best:
AUC: 0.687

For the first-time PySpark ML user this can come out as a bit disappointing. They were probably expecting the PySpark ML engine to do more "magic" like automatically selecting the best features, tuning hyperparameters, or trying multiple models to find the absolute best fit all on its own. But it's still just the classic logistics regression, which surprisingly also falls under the umbrella of ML.

Train dataset (80% of the data)

Actual	Prediction	Count
0	0	89285
0	1	191
1	0	6407
1	1	252
Total		96135

Test dataset (20% of the data)

Actual	Prediction	Count
0	0	22384
0	1	52
1	0	1618
1	1	80
Total		24134

---

Python language reminders:

Mutable objects can be changed after they are created. You can modify their contents without creating a new object. When you change a mutable object, the object’s identity (its memory address) stays the same.

Immutable objects cannot be changed after they are created. Any "change" to an immutable object creates a new object instead. When you try to modify them, you actually get a new object with the change applied.

Type	Mutable?
int	No
float	No
bool	No
str	No
tuple	No
frozenset	No
bytes	No
list	Yes
set	Yes
dict	Yes
bytearray	Yes

Type	Explanation
int	Integer numbers, whole numbers without a decimal point (e.g., 42, -7).
float	Floating-point numbers, numbers with decimals (e.g., 3.14, -0.001).
complex	Complex numbers with real and imaginary part; a + bj where j is the imaginary unit.
bool	Boolean type representing truth values: True or False.
str	Strings, sequences of characters used for text (e.g., "hello").
tuple	Immutable ordered sequence of elements (can be mixed types), e.g., (1, 2, 3).
frozenset	Immutable version of a set, an unordered collection of unique elements that cannot be changed.
bytes	Immutable sequence of bytes, used for binary data (e.g., file content, network data).
list	Mutable ordered sequence of elements, can be changed by adding, removing, or modifying items.
set	Mutable unordered collection of unique elements, useful for fast membership tests and operations.
dict	Mutable collection of key-value pairs (a mapping), e.g., {"a": 1, "b": 2}.
bytearray	Mutable sequence of bytes, similar to bytes but can be modified.