Muhammad Umair Nasir

PhD Student and Sessional Lecturer @ University of the Witwatersrand
muhammad.nasir@wits.ac.za

I am a PhD student at the University of the Witwatersrand, Johannesburg, South Africa. I am a member of the RAIL Lab research group and NYU Game Innovation Lab research group. I am honoured to be advised by Dr. Steven James and Prof. Julian Togelius. I am also a sessional lecturer at the University of the Witwatersrand. My main research interest is in the intersection of AI and games, specifically Open-Ended Learning and Procedural Content Generation.

Experience

University of the Witwatersrand

Sessional Lecturer

Johannesburg, South Africa
Jul 2024 - Present

  • Teaching "Data Visualisation and Exploration" to 119 Honours and Masters students.

Axon Corporate Services

Data Scientist

Johannesburg, South Africa
Feb 2023 - Apr 2024

  • Building data pipeline using Spark.
  • Building intelligent internal threshold checker to prevent showing wrong currency values to client.
  • Using Machine Learning tools to predict the causes of currency value threshold breaches.
  • Researching and developing LLM-based time series model for currency prediction.

Ominor AI

Associate Machine Learning Engineer

Johannesburg, South Africa
Dec 2021 - Dec 2022

  • Created end-to-end video-based real-time diabetes detection system.
  • Worked on BigQuery for data warehousing.
  • Used GCP to deploy machine learning algorithms.
  • Used GPT-3 to create a specific-niche article writer for a client.
  • Optimized for cost and inference to deploy large language models.
  • Worked with health-tech AI client to automate a diabetes coaching system to create a recommender system to recommend what interventions need to be applied.
  • Worked with logistics client to forecast total liter consumption by their fleet through fbprophet.

Aureks

Data Scientist Intern

Remote
May 2021 - Nov 2021

  • Worked on machine learning pipeline that uses Kmeans clustering to maintain semantically similar keywords in clusters.
  • Used spaCy for finding semantic similarities between words. Word embedding vectors were used to find the distance between the new keyword and the cluster.
  • Applied PCA for dimension reduction of word embedding vectors.
  • Passing the new words to the Fbprophet time series framework to predict Cost-Per-Click of the keywords.
  • Maintaining Data pipeline through MongoDB and pymongo.

Reliance Engineering Company

Site Engineer

Multan, Pakistan
Dec 2018 - Nov 2020

  • Manage onsite team installing high-voltage transmission lines.
  • Coordinate with external contractors to manage supply of the project.

ZTE Corporation

Electrical Engineer

Aug 2016 - Nov 2018

  • Supervise the installation of electrical components of a government-owned broadcast station.

Freelance Experience

Procedural Room Generation For Metavoidal

(https://yellowlabgames.itch.io/metavoidal)

Yellow Lab Games, South Africa
March 2023

  • Created a pipeline for room generation through GPT-3's OpenAI API.
    This also led to the research paper "Procedural PCG Through Large Language Models".

Wrote Chapters For The Book: "Data and AI"

(https://edgeeducation.com)

Edge Education, South Africa
December 2023

  • I am the lead author in the book named "Data and AI" which is targeted on 1st year undergraduate students.
  • I wrote four chapters for the book.

Built RAG Pipeline with Aleph Alpha's LLM

(https://ritza.co)

Ritza, South Africa
January 2024

Contributor to open-source web application: Baobab

(https://github.com/deep-learning-indaba/Baobab)

Baobab
Feb 2024 - present

Publications

Contact Me

You can contact me at my email: muhammad.nasir@wits.ac.za