manzan.dev

What have I been up to during the last ~2.5 years?

❯ Work

Note: I started a new job in 11/2020 👀

Previously, I worked on developing a cloud native banking platform on Google Cloud Platform, from scratch.

We built components of a data platform back-end, for example tools for infrastructure provision/config using Terraform and Python, schema management/validation using Protobuf, Python and Go, or data quality validation using dbt (data build tool), SQL and Argo for orchestration. CI/CD was done with Docker, Jenkins and Kubernetes. Data pipelines consisted of Pub/Sub, Dataflow (Apache Beam) and BigQuery.

Further highlights:

• Google Cloud Certified Associate Cloud Engineer (02/2020) – learnt how to deploy applications, monitor operations, and manage enterprise solutions based on Google Cloud
• Udacity Machine Learning Engineer Nanodegree (06/2020) – learnt how to build, train, and deploy Machine Learning models to a production environment using Amazon SageMaker
• Data & Python Software Engineer for Machine Learning applications within LBG ML lab
• Instructor for Software Engineering training of 2019 Data Science graduate class (intro to bash & git)

❯ Fun

I enjoy learning new technologies and working on ideas in my spare time. Below you can find some recent mini projects:

• manzan.live – online radio station/API built with julvo using FastAPI, TypeScript, React & Firebase
• Scraper for Resident Advisor using Scrapy
• Simple Python UDP server using asyncio, plus Dockerfile, Makefile and unittests
• Social distancing measures of London tube stations using Google Maps Places API
• Q&A engine hackathon project using Flask & BERT model
• Dog breed classifier as part of ML Engineer Nanodegree using PyTorch ResNet model

I've also been an early contributor to Whale 🐳, an open-source data discovery, documentation, and quality engine created by Robert Yi and Joseph Moon. Read more about it here.

❯ Research

For my M.Sc. thesis I've collaborated with researchers of the IOTA Foundation in Berlin and Tel Aviv in cooperation with the IC Centre for Cryptocurrency Research and Engineering and Outlier Ventures.

I developed a simulation framework in Python and published a short paper summarising the main findings together with Tom Waite and Dominik Harz in ACM SIGMETRICS Performance Evaluation Review in 01/2019. The abstract is as follows:

Scalability of distributed ledgers is a key adoption factor. As an alternative to blockchain-based protocols, directed acyclic graph (DAG) protocols are proposed with the intention to allow a higher volume of transactions to be processed. However, there is still limited understanding of the behaviour and security considerations of DAG-based systems. We present an asynchronous, continuous time, and multi-agent simulation framework for DAG-based cryptocurrencies. We model honest and semi-honest actors in the system to analyse the behaviour of one specific cryptocurrency, IOTA. Our simulations show that the agents that have low latency and a high connection degree have a higher probability of having their transactions accepted in the network with honest and semi-honest strategies. Last, the simulator is built with extensibility in mind. We are in the process of implementing SPECTRE as well as including malicious agents.

You can download the paper here, and find my full thesis here.