m
m_irfan_eng

Muhammad Irfan

@m_irfan_eng

Data Scientist AI ML OCR Gen AI Solutions

Pakistan
Anglais, Ourdou, Punjabi
Certaines informations sont présentées en anglais.
À propos de moi
Stop paying for manual work and costly data errors. I am an expert Python Automation Engineer (ex-AI4LYF) dedicated to building highly robust and scalable data pipelines that cut administrative costs by up to 40%. My Core Value: I transform messy, unstructured data (scanned forms, receipts, clinical reports, sensor logs) into clean, usable formats for immediate analysis. Specialized Expertise: OCR Automation, Data Visualization & Dashboards, PDF-to-Excel script, biomedical classification model... Plus d’infos

Compétences

m
m_irfan_eng
Muhammad Irfan
hors ligne • 
Temps de réponse moyen de 1 heure

Voir mes services

Automatisations
I will automate PDF and document data extraction using python ocr
De30 $US / projetPlus de détails

Expérience professionnelle

AI Engineer

AI for lyfe • Temps plein

Jun 2021 - Oct 2024 • 3 yrs 4 mos

Led end‑to‑end data processing pipelines for wearable sensors and mobile microphones, performing large‑scale cleaning, feature extraction, and behavioral/physiological signal analysis for health insights. Built and optimized a Pix2Pix GAN with MAE, VGG, content, and classifier‑based losses, improving reconstruction quality by 50%. Developed a CNN‑based multiclass, multilabel classifier trained on hybrid real + GAN‑generated datasets, achieving 92% accuracy and demonstrating synthetic data reliability. Engineered an interactive biomarker‑monitoring dashboard in Python Dash, accelerating clinical reporting by 30% and enabling real‑time visualization of key health indicators. Automated OCR workflows using OpenCV, EasyOCR, and AWS Textract, reducing manual data‑entry time by 40% and increasing accuracy. Designed XGBoost and ExtraTrees models for biomedical audio classification across four clinical conditions, integrating demographic metadata for improved robustness. Led data‑infrastructure efforts, designing a scalable collection framework (20K+ validated samples), backend systems, validation pipelines, and team training to ensure safe and compliant data handling.