---
title: Words
subtitle: "A collection of words I didn't know"
bullets: 
  [
    {
      item_colour: "#1EA808",
      value: "B" 
    },
    {
      item_colour: "#FFFFFF",
      font_colour: "black",
      value: "3" 
    },
  ]
date: "2023-11-21"
slug: "word-definitions"
skip: true
---

I like to keep tabs on words I come across that I may, not know, or sort of know in the context of the sentence but not understand the word deeply.


2023-11-21-book-definitions.mdx

A collection of words I didn't know

Words

---
title: My Website
subtitle: A personal portfolio and other things
tags: [aws, frontend, design]
item_colour: "#fccc0a"
bullets: 
  [
    {
      item_colour: "#b933ad",
      value: "F" ,
      font_colour: "white"
    },
    {
      item_colour: "#ff6319",
      value: "A" ,
      font_colour: "white"
    },
    {
      item_colour: "#fccc0a",
      value: "D" ,
      font_colour: "white"
    },
  ]
item_font: "white"
date: "2023-07-24"
slug: "personal-website"
skip: true
---

import WebsiteFonts from "../components/mdx/WebsiteFonts";

I decided to build a website to document my work as well as having a place were I can share my writing about topics I find interesting.

![](../images/iphone-mockup-website.png)

## Plumbing

I used [GitHub pages](https://pages.github.com/) to host the website and a Jekyll [theme](https://jekyllthemes.io/theme/made-portfolio-jekyll-theme) as a base.

## Typography

I love typography and spent a lot of time choosing the right fonts that express my design inclinations.

<>
<div class="grid">
  <div>
    <h1 class="font__grotesk">Founders Grotesk</h1>
  </div>
  <div class="chars__box">
    <h2 class="chars__grotesk" >
      abcdefghijklmnopqrstuvwxyz
    </h2>
    <h2 class="chars__grotesk">ABCDEFGHIJKLMNOPQRSTUVWXYZ</h2>
    <h2 class="chars__grotesk" >
      1234567890
    </h2>
  </div>
  <div class="font__desc">
    <p>
      Founders Grotesk is a beautiful font from the folks at{" "}
      <a href="https://klim.co.nz/">Klim Foundry</a>.
    </p>
    <p>
      The type is used throughout the site in document headings and as the logo
      for the footer and header (my name). It's definitely money well spent and
      looks gorgeous on the web on both mobile and laptop.
    </p>
    <p>
      A large bold headline really brings out this type and is how I have used
      it throughout this site.
    </p>
    <p>
      More information on the design can be found{" "}
      <a href="https://klim.co.nz/blog/founders-grotesk-design-information/">here</a>
      if you are interested.
    </p>
  </div>
  <div>
    <h1 class="font__mono">Berkeley Mono</h1>
  </div>
  <div class="chars__box">
    <h2 class="chars__mono">
      WORLD LEADING PRODUCT INNOVATION Makita® Corporation was founded in 1915
      as an electric motor sales and repair company. Today, as a global brand in
      over 40 countries, Makita
    </h2>
  </div>
  <div class="font__desc">
    <p>
      Berkeley Mono is one of my favorite releases this year.
      <a href="https://berkeleygraphics.com/">Berkeley Graphics </a> 
      produced the
      type and their philosophy — Engineering graphics for professionals —
      really shines. I love everything about Berkeley Graphics, especially their{" "}
      <a href="https://twitter.com/berkeleygfx">Twitter</a>.
    </p>
    <p>
      The title of blog items, projects and the grid on the blog page use
      Berkeley Mono. It makes for a clean, readable type for content that is
      usually programming related. Having titles such as:
      <h2 class="mono__example">
        <span></span>TIL: bitwise ǀ and & on open flags in UNIX
      </h2>
    </p>
    <p>lend themselves to a crisp display using Berkeley Mono.</p>
    <p>
      I use the type in my IDE and the all code blocks here. It looks clean af.
    </p>
  </div>
  <div class="code__block"></div>
  <div>
    <h1 class="font__london">London Underground</h1>
  </div>
  <div class="chars__box departures">
    <div class="chars__london">
      <div class="departure__item">
        <div>4</div>
        <div>Bank</div>
        <div class="eta">2 min</div>
      </div>
    </div>
    <div class="chars__london">
      <div class="departure__item">
        <div>3</div>
        <div>Clapham Common</div>
        <div class="eta">3 min</div>
      </div>
    </div>
    <div class="chars__london">
      <div class="departure__item">
        <div>1</div>
        <div>Victora</div>
        <div class="eta">10 min</div>
      </div>
    </div>
  </div>
  <div class="font__desc">
    <p>
      I moved to London in 2022. I catch the tube a lot. Someone dedicated a{" "}
      <a href="https://github.com/petykowski/London-Underground-Dot-Matrix-Typeface">whole project</a>
      which I found on
      <a href="https://news.ycombinator.com/item?id=36367667">HN</a> emulating
      what we see on the LED screen when waiting the short time for the tube to
      arrive. It's awesome.
    </p>
    <p>
      This is a homage to a new home for me and ends up in a couple of locations
      on the site. I use it for the dates of my posts and projects as well as
      the previous and next post text at the end of a blog post.
      <h2 >
        13 DECEMBER 2021{" "}
      </h2>
      <h2 >
        24 JULY 2023{" "}
      </h2>
    </p>
  </div>
  <div>
    <h1 class="font__mono_tronica">Tronica Mono</h1>
  </div>
  <div class="chars__box">
    <div class="chars__mono_tronica">
      <div class="grep__item">grep(1)</div>
      <div>NAME</div>
      <div class="grep__name">
        tgrep, egrep, fgrep, rgrep, bzgrep, bzegrep, bzfgrep, zgrep, zegrep,
        zfgrep – file pattern searched
      </div>
    </div>
  </div>
  <div class="font__desc">
    <p>
      Tronica Mono is an awesome pixelated terminal like typeface. I love
      anything pixelated. Tronica makes it's way around the site in a few areas.
      In the year part of the date in my footer, the numbers that order my
      projects, small heading on the home page and lastly in the About Me page.
    </p>
  </div>
</div>

## Colours

Most the colours chosen are inspired by various metropolitan subway stations. For instance the New York Subway.

<p align="center">
  <img src="../images/mta-lines.jpeg"></img>
</p>

Of course I have ran out of colours to use from the New York Subway. The tube to the rescue.

<p align="center">
  <img ></img>
</p>
</>

Eventually I'll have to start using the colours of Skittles or M&Ms.

## Images

I always try to create my own meaningful images that relate to what I am writing about.

![](../images/projects/../gallery/gba-lg.png)

For instance this is a screen grab from a short-lived game based on an awful movie (Toxic Avenger) that is popular with the B-movie crowd.

I used a variation of this in my page about my bad movie project. I made it in Adobe Photoshop using a Nintendo Game Boy Advance mockup.

![](../images/projects/kafka-docker-project/notebook_feature.png)

Another example, for my Bitcoin pipeline project. This features a container (docker) mockup in a notebook, with some awesome marker assets (kafka) I found on Behance. This was also made on Photoshop. I enjoy the process of creating unique graphics to make my posts stand out.

## Inspiration

I find inspiration from unlikely places. Just observing things that I see when I walk around or just getting a general _feel_ for what I want to achieve visually works too.

But I couldn't do without these awesome resources:

- [Pinterest](https://www.pinterest.co.uk/)
- [Behance](https://www.behance.net/)
- [Dribble](https://dribbble.com/shots)
- [awwwards](https://www.awwwards.com/)

I think a love of pop culture and culture in general most definitely informs my style and general inclinations towards certain visuals and styles.

<WebsiteFonts />


2023-07-24-personal-website.mdx

A personal portfolio and other things

My Website

---
title: Alfred Hackernews 
subtitle: Read the latest from Hackernews from within Alfred
bullets:   
  [
    {
      item_colour: "#0039a6",
      value: "P" 
    },
    {
      item_colour: "#a7a9ac",
      value: "A" 
    },
  ]
project_url: "https://github.com/kostyafarber/alfred-hackernews"
item_colour: "#ff6319"
date: "2023-06-03"
description: "Serving up the latest news from hackers right in Alfred"
slug: "alfred-hackernews"
---

Serving up the latest news from hackers right in Alfred

2023-06-03-alfred-hackernews.mdx

Read the latest from Hackernews from within Alfred

Alfred Hackernews

---
title: Alfred Notable
subtitle: Make and search notes in Notable from Alfred
bullets: 
  [
    {
      item_colour: "#0039a6",
      value: "P" 
    },
    {
      item_colour: "#a7a9ac",
      value: "A" 
    },
  ]
item_colour: "#a7a9ac"
project_url: "https://alfred.app/workflows/kostyafarber/notable/"
description: "Alfred workflow for Notable"
date: "2023-06-03"
slug: "alfred-notable"
---


2023-06-03-alfred-notable.mdx

Make and search notes in Notable from Alfred

Alfred Notable

---
title: Vice City API 
subtitle: "Find your favorite radio stations and music in vice city"
layout: projects
bullets:
  [
    {
      item_colour: "#00add0",
      value: "B" 
    },
    {
      item_colour: "#0039a6",
      value: "P" 
    },
    {
      item_colour: "#996633",
      value: "D" 
    }
  ]
description: "Find your favorite radio stations and music in vice city"
project_url: "https://github.com/kostyafarber/vice-city-api"
date: "2023-06-03"
slug: "vice-city-api"
---

# some stuff

hi

Find your favorite radio stations and music in vice city

2023-06-03-vice-city-api.mdx

Vice City API

---
title: Collection of Design Related Items
subtitle: "A collection of design related items I've produced over the years"
bullets: 
  [
    {
      item_colour: "#fccc0a",
      value: "D" 
    },
    {
      item_colour: "#b933ad",
      value: "F" 
    },
  ]

date: "2023-05-06"
description: "Some designs I've produced over the years"
slug: "front-end-gallery"
skip: true
---


Some designs I've produced over the years

2023-05-06-front-end-gallery.mdx

A collection of design related items I've produced over the years

Collection of Design Related Items

---
title: Bitcoin Real-Time Streaming Data Pipeline
subtitle: "A DevOps project utilising Docker, Kafka and Microservice Architecture Principles"
date: "2022-08-06"
slug: "crypto-kafka-screaming"
bullets:
  [
    { item_colour: "#996633", value: "D", tag: "data" },
    { item_colour: "#0039a6", value: "P", tag: "python" },
  ]
---

import { Figure } from "../components/mdx/Figure";

I originally intended to create an algorithmic trading system based on limit order book (LOB) data. I wanted to trade cryptocurrency **perpetual** futures.

> The Deribit Perpetual is a derivative product similar to a future, however, without an expiry date. The perpetual contract features funding payments. These payments have been introduced to keep the perpetual contract price as close as possible to the underlying crypto price - the Deribit BTC Index.

I also wanted to choose a platform that provides an API where I could access live LOB data. I chose [Deribit](https://www.deribit.com/) as they meet both my needs.

<Figure caption="Deribit Logo">
  <img
    src="../images/projects/kafka-streaming/deribit-logo.png"
    alt="Deribit Logo"
    width="400"
    style={{ margin: "auto" }}
  />
</Figure>

## Python Websocket

Deribit provides three different interfaces to access the API:

- JSON-RPC over Websocket
- JSON-RPC over HTTP
- FIX (Financial Information eXchange)

I decided to use JSON-RPC over Websocket. JSON-RPC is a light-weight remote procedure call (RPC) protocol. The JSON-RPC specification defines the data structures that are used for the messages that are exchanged between client and server, as well as the rules around their processing. JSON-RPC uses JSON (RFC 4627) as data format.

```python caption="An example of a request message"
{
    "jsonrpc": "2.0",
    "id": 8066,
    "method": "public/ticker",
    "params": {
        "instrument": "BTC-24AUG18-6500-P"
    }
}
```

I used [this](https://github.com/ElliotP123/Deribit-API_Authentication-Examples/blob/master/JSON-RPC_Websockets/JSON-RPC_Websockets-AuthenticationExample.py) code as a base to create my websocket client in Python to access the live streaming data for the perpetual futures. It handles the authentication and subsequent subscription of the required data.

## Subscription Management

The subscribe method can be used to subscribe to one or more channels. In most cases the channel name is constructed from a couple of elements. This makes it possible to specify exactly which information is required, and/or the frequency or aggregation level. These elements are considered parameters for the subscription.

For example, when subscribing to the channel `book.BTC-27JUL18.10.20.100ms`, the element `BTC-27JUL18` specifies that the name of the instrument (see naming), 10 means that the results should be grouped to that precision, etc.

## Streaming Order Book Data

For my purporses I subscribe to the `book.{instrument_name}.{group}.{depth}.{interval}`, which notifies the subscriber about changes to the order book for a certain instrument.


```python caption="An example notification for Ethereum (ETH)."
{
  "params" : {
    "data" : {
      "timestamp" : 1554375447971,
      "instrument_name" : "ETH-PERPETUAL",
      "change_id" : 109615,
      "bids" : [
        [
          160,
          40
        ]
      ],
      "asks" : [
        [
          161,
          20
        ]
      ]
    },
    "channel" : "book.ETH-PERPETUAL.100.1.100ms"
  },
  "method" : "subscription",
  "jsonrpc" : "2.0"
}
```

## Limit Order Book

The objective was to obtain an indicator that could point to the state of the market at a certain point in time. An **order book** is the list of orders that a trading venue uses to record the interest of buyers and sellers in a particular financial instrument.

<Figure caption="Limit Order Book Diagram">
  <img
    src="../images/projects/kafka-streaming/lob-diagram.png"
    width="500"
    style={{ margin: "auto" }}
  />
</Figure>

## Order Book Imbalance

The Order Book Imbalance is the difference between the best bid and best ask quantity quotes divided by its sum. Such that,

```math caption="dsa"
Imb*{t} = \frac{V*{t}^b - V*{t}^a}{V*{t}^b + V\_{t}^a}
```

The Order Book Imbalance measures whether the limit order book is buy or sell heavy. In fact, it is a good predictor of price direction.

## Designing the Architecture

Having all the groundwork laid in place I went about deciding on how I wanted to design this system. To make things simpler and get my hands dirty with the current trend of microservices, I wanted to use Docker and Kafka to build a simple and scalable ETL pipeline.

## Docker

Docker is a valuable tool for developers to ship and deploy their apps in an easy and accessible way. We can create _containers_ that are self-contained environments that house our applications. Taking this a step further, we can decouple our system and build a microservices architecture by having our services in their own docker containers. I will use Docker to containerize a few services used for the data pipeline built using Kafka.

## Kafka

Apache Kafka is an open-source distributed event streaming platform used by thousands of companies for high-performance data pipelines, streaming analytics, data integration, and mission-critical applications.

It uses a publisher and subscriber architecture, both of which are decoupled from the core functionality of Kafka. Apache Kafka has an API in Python called `kafka-python`. Using `kakfa-python` I am able to build the data pipeline using Python.

## Data Pipeline

The image below demonstrates the data pipeline utilising Docker and Kafka. We have seperate containers containing each microservice and Kafka itself. The Deribit producer container connects to the Derbit API and streams JSON records and publishes them to Kafka in a channel called _prices_.

Seperately, the consumer connects to the _prices_ topic and transforms the raw JSON into order flow imbalance and mid-price variables.

## Microservices

Each service is unaware of the other but communicate through docker networking. By using Docker and Kafka we are able to create a scalable and fault tolerant architecture were each service is decoupled from one another. This is at the heart of modern day software engineering and enables large distributed systems to work efficiently.

## Wrapping Up

The purporse of this project was to get my hands dirty with real-time streaming data and containizeration technologies.
I was able to write code that consumed an API delivering JSON over websocket.
I designed a small distributed (capable) data pipeline using Kafka and containerized my services to adhere to microservice principles.
I was able to transform raw data to obtain useful Market Microstructure variables such as order flow imbalance and the mid-price.

Make sure to check out my GitHub for the source code and how to run it yourself!


2022-08-06-crpyto-perpetual-futures-kafka-streaming.mdx

A DevOps project utilising Docker, Kafka and Microservice Architecture Principles

Bitcoin Real-Time Streaming Data Pipeline

---
title: Car Sales Scraper
bullets:
  [
    {
      item_colour: "#ff6319",
      value: "A" ,
      tag: "aws",
      font_colour: "white",
    },
    {
      item_colour: "#0039a6",
      value: "P",
      tag: "python",
      font_colour: "white",
    },
    {
      item_colour: "#996633",
      value: "D", 
      tag: "data",
      font_colour: "white",
    },
  ]

date: "2022-07-03"
subtitle: "Web scraper that helped sell my car using a data-driven approach"
slug: "car-sales-scraper"
---

import { Figure } from "../components/mdx/Figure"

In mid-2022 I made the move to relocate to London. An inevitable part of moving overseas is getting rid of any belongings you don't need to take with you. One of these belongings was my _2013 Mazda 3 Sedan_. A fundamental question that needs to be answered is:

> How much is my car worth on the second hand market?

I've found myself an interesting problem that I needed to solve to help me get a fair price for my car. I decided to create an application that would give me a data-driven answer to my problem.

## How Did I Do It?


A popular site, [carsales.com.au](https://www.carsales.com.au/) allows users to sell their second-hand cars. This acts as a marketplace for buyers and sellers. 

<Figure caption="car sales website">
  <img src="../images/projects/car-sales/car-sales-mockup.png" /> 
</Figure>

I searched for cars that fit the description of my Mazda. From here I could grab all the prices of the cars on the market (on carsales.com.au) and base my price of this. I used [BeautifulSoup](https://beautiful-soup-4.readthedocs.io/en/latest/) to scrape the html of the page and find the sales prices.


## Code

Using Google Chrome's inspect, I isolated the correct class, aptly named `title` in the `h1` tags. I used a class to implement my scraper called `CarSalesScraper`. It performs all the scraping tasks and data cleansing needed for ingestion into my database. You can find the source code in the GitHub link above.

## AWS
I employed Amazon Web Services (AWS) to manage my database solution, daily execution of the code and email notifications.

<Figure caption="aws services used">
  <img src="../images/projects/car-sales/tech-stack.png" />
</Figure>

- **AWS Lambda** managed the daily execution side.
- **AWS DynamoDB** was used to store the average, median and raw sales price
- Finally **AWS Simple Notification Service (SNS)** was used to send email notifications when my DynamoDB was written to.

## Challenges

One challenge I ran into was using packages that aren't available in the off the shelf Lambda. To make matters slightly more difficult my solution locally is on a **_Mac OS X_**. Lambda runs on **_Linux_**. So I couldn't simply package my implementation of BeautifulSoup onto a lambda layer as it's coming from a Mac.

## Solution
After some research online I found an easy solution within the AWS ecosystem. Cloud9.

<Figure caption="cloud9 and linux logos">
  <img src="../images/projects/car-sales/cloud9-linux.png" />
</Figure>

> AWS Cloud9 is a cloud-based integrated development environment (IDE) that lets you write, run, and debug your code with just a owser.

We can spin up this environment run on top of Linux and create a zip `deployment.zip` and use it to install packages into this zip.

```sh caption="installing pip packages into a target directory"
pip3 install -t . your_packages
```

Adding this zip to Lambda as a layer allows the managed service to access our package and our script can execute.

## Outcome

Using this solution I was able to get a gauge on the market. Every morning I got updates on the average, median and raw prices. I was able to adjust the prices accordingly day to day. Eventually I successfully sold my car for a good price. Being able to have data to inform my decision put me at ease and allowed me to set prices confidently. Using the cloud made creating this application a eeze and the joy from using technology to help others and myself is priceless!

source: [https://github.com/kostyafarber/car-sales](https://github.com/kostyafarber/car-sales)


2022-07-03-car-sales-scraper.mdx

Web scraper that helped sell my car using a data-driven approach

Car Sales Scraper

---
title: B-Movie Text Generator
subtitle: What if a robot could write b-movies?
bullets:
  [
    {
      item_colour: "#0039a6",
      value: "P", 
      tag: "Python",
      font_colour: "white"
    },
    {
      item_colour: "#996633",
      value: "D", 
      tag: "data",
      font_colour: "white"
    },
  ]
date: "2021-01-29"
slug: 'bad-movies'
---

import { Figure } from "../components/mdx/Figure"

## Motivation

One of my fond memories of a bygone era was visiting the local video store, the now ancient ***Video Ezy***, to pick up a movie with my friends. We would go
to each others houses and have a movie marathon and watch all types of films. Some good, some bad.

<Figure caption="video ezy logo">
  <img className="rounded-xl" src="../images/projects/bad-movies/video-ezy.jpg" />
</Figure>

Over the years we all naturally became avid *cinephiles*. Analysing films while watching them became a favourite pastime and
surprisingly, the bad films had plenty of (hilarious) things to analyse. What followed was uncovering a giant underground cult following of all things
bad.

One particularly brilliant medium that explored this sub-genre of movie, was the podcast *How Did This Get Made*.

> How Did This Get Made? is a podcast on the Earwolf network. It is hosted by Paul Scheer, June Diane Raphael and Jason Mantzoukas. Each episode, which typically has a different guest, features the deconstruction and mockery of outlandish and bad films.

After listening to this podcast religiously, me and my friends decided, why not create our own corpus of bad movies. So the bad
movie club is born, and 200+ movies later a rather interesting (to me at least) dataset is produced.

I've always had the idea to create a data science project from our long and silly list, so here it is.

source: [https://github.com/kostyafarber/b-movie-text-generator](https://github.com/kostyafarber/b-movie-text-generator)


2021-01-29-bad-movies.mdx

What if a robot could write b-movies?

B-Movie Text Generator

---
title: Sydney Property Prices Dashboard
subtitle: "Sydney property market dashboard across different suburbs using a dataset from Kaggle."
bullets: 
  [
    {
      item_colour: "#b933ad",
      value: "F", 
      tag: "frontend",
      font_colour: "white"
    },
    {
      item_colour: "#0039a6",
      value: "P", 
      tag: "python",
      font_colour: "white"
    },
    {
      item_colour: "#996633",
      value: "D",
      tag: "data",
      font_colour: "white"
    },
  ]
date: "2021-01-28"
slug: 'sydney-house-prices'
---

## The Motivation

As the old adage goes, for pragmatists anyway, you learn by doing. This especially applies to learning programming. 
So I set off into the virtual world to find real world data that wasn't curated to make the coders life easier. 

I first scoured [Kaggle](https://www.kaggle.com/) for data I could use to build an interesting project. My first instinct
was to find data that was close to me in terms of proximity.  

>Kaggle allows users to find and publish data sets, explore and build models in a web-based data-science environment, work with other data scientists and machine learning engineers, and enter competitions to solve data science challenges.

A simple search of *'Sydney'* into the search bar yielded 
a few results, but being the typical Sydney resident, data of property prices seemed like an obvious option.

## The Result
- A fully custom built interactive dashboard programmed from scratch using **Dash Plotly**, a Python framework for building web analytic applications. 
- The front end application is deployed on **Heroku**, a cloud platform that makes it easy to deploy and manage Flask applications. 
- The data was obtained from Kaggle, a dataset with over 100,000 entries scraped from realestate.com.
- Data wrangling, Exploratory Data Analysis and Data Visualisations were performed using Python and documented in both static and executable (using Binder) Jupyter Notebook for reproducibility.
- All code and project details are open-source and shared on GitHub. 

source: [https://github.com/kostyafarber/sydneyhouseprices-dashboard](https://github.com/kostyafarber/sydneyhouseprices-dashboard)

Software —

engineer.

always

Alfred Hackernews

Alfred Notable

Vice City API

Bitcoin Real-Time Streaming Data Pipeline

Car Sales Scraper

B-Movie Text Generator

Sydney Property Prices Dashboard

work.

Aviva Zero

Aviva

KPMG