Secure Data Vault
Client: Selstan
Role: Machine Learning Engineer
Date: Mar 2025 – Apr 2025
Built a prototype data redaction API to detect and replace sensitive information in real-time, enabling privacy-first handling of sensitive text data. Fine-tuned a BERT NER model and integrated it with H2O.ai’s DeBERTa PII detector. Deployed the pipeline via Flask as a modular API.
🛠️ Tech Stack: Python, Flask, Hugging Face Transformers, BERT, DeBERTa
“It is with confidence that I recommend Umair. He possesses a dedicated commitment to assigned tasks and a strong work ethic. His technical skills are highly recommendable, including his proficiency as a Machine Learning Engineer. We would be pleased to have him collaborate with our engineering team again."
— Vitalis Omambia, Strategic Lead
AI Chatbot Assistant
Client: Atlantis AI
Role: AI Developer
Date: Jan 2025 – Mar 2025
Built an AI-powered assistant chatbot to answer user queries about credit card rewards using Retrieval-Augmented Generation (RAG). Implemented vector-based retrieval with LlamaIndex and Sentence Transformers, and generated responses using TinyLlama via Hugging Face Transformers.
🛠️ Tech Stack: Python, LlamaIndex, TinyLlama, Sentence-Transformers, Hugging Face Transformers, Selenium
“Umair explored the project in depth and was creative in his code debugging and creation. Umair is a detail oriented student and excellent for any data pipeline related project."
— Alexander Moreno-Ip, Founder
Website Optimization
Client: I Challenge Diabetes
Role: Website Developer
Date: Feb 2024 – Apr 2024
Optimized a nonprofit organization’s site to improve search visibility, performance, and long-term maintainability. Increased SEO performance by 140% and reduced total site size by 880 MB. Authored a guide to enable the team to independently maintain the site.
🛠️ Tech Stack: WordPress, HTML/CSS
“Your collaboration, dedication, and effort have been outstanding, leading the overall success of our project. Keep up the excellent work and thank you for your continued efforts! Well done!"
— Imran Nathani, Operations Manager