Zain Ul Abideen Email & Phone Number

Lewes, Delaware, United States

London Area, United Kingdom

1) Implemented a self-optimizing multimodal pipeline with RAG, Agentic workflow, and open-source AI using LLM-as-a-Judge and Mixture of Agents.2) Managed 30+ GPUs for multi-node inference of the entire multimodal pipeline consisting of LLama-3.1 70B, Phi-3-medium-128k-instruct, Llava-next-8b, and SDXL-Lightning.

San Francisco Bay Area

1) Pre-trained a 500M SLM from scratch on a carefully curated high-quality 15B tokens synthetic dataset. Created the entire training and evaluation pipeline along with managing training on 8xA100s.2) Experimenting and Contributing to various LLM OS projects like Mixture of experts, Hybrid MOE, Bitnet, Mixture of Depths, LaserQlora, and Daser.3) Heading the "LLM.pth" opensource project. The main objective of this library/codebase is to provide implementations of various Autoregressive models, Research papers, and techniques in Pytorch.

Engineered an end-to-end pipeline for time series forecasting of bandwidth needs to maximize satellite communication using LightGBM, which consistently outperformed human baseline strategies by 2-4%.Annotated images and fine-tuned Yolov5 model for the detection of paddy rice for a research publication.Built RAG pipeline for Real Estate company documents with Ray and Langchain. Increased the overall productivity workflow of employees by 5-8%.Developed a Machine Learning pipeline for an online retail organization and applied Agglomerative clustering to get valuable insights leading to a 5% increase in sales. This whole pipeline was built on AWS.Improved classification accuracy by 4% for the existing churn prediction model. Deployed machine learning models on AWS using Kubeflow. Built Microservices with FastAPI and deployed them on AWS EKS.

Built Exar, the company's first-ever fully AI-reliant Exercise engine based on ConvLSTM, and Blazepose for 10 schools in USA and India. Aggregated and prepped 400 videos from 4 data sources. Deployed the flask server on AWS EC2 and used DynamoDB for intermediary outputs.Developed a game pose classifier with Tensorflow and Movenet, and collaborated with a team of C# engineers to deploy it on Unity. The product had 1000 daily active users.Built Diagno, a multi-stage computer vision model to automate the process of disease detection from CBCT dental scans. The product was used by two dental clinics with 300 monthly patients.

Worked on the research of Transformer based models in the field of Image Segmentation and Time series forecasting for the monitoring of traffic on multi-lane highways.Developed the entire system design and ML pipelines for monitoring traffic on AWS with Kinesis, Sagemaker, Lambda, and Quicksight.

Bachelor'S Degree, Software Engineering

Education record