Saqib Ahmed Qureshi

Experience

• 

  TeamKART Motorsports

  Feb 2019 - Jan 2020

  • Suspensions , Braking and Steering Member

    Aug 2019 - Jan 2020

  • Corporate and Public Relations Member

    Feb 2019 - Jan 2020
• 

  HeadStart RMSOEE Student Cell, IIT Kharagpur

  May 2020 - Jun 2020

  Data Science Intern

  Name of the startup - DAWN• Collected and analysed the data of various crops across all over India which includes their prices, the season of cultivation, place suitable for their farming and other factors which are essential for increasing the yield productivity.• Pre-processed the data in the form which is suitable to visualise it and understand it better in python.• Prepared data for predictive analysis of prices of crops through LSTM model.

• 

  Effat University

  Jun 2020 - Aug 2020

  Summer Research Intern

  COVID-19 detection from X-ray images using Deep LearningInstructor - Prof. Abdulhamit Subasi• 11 pre-trained models (VGG16, VGG19, ResNet, ResNet50, InceptionV3, InceptionResNetV2, MobileNet, MobileNetV2, DenseNet121, DenseNet169, Xception) were used to build a categorical classifier which classifies among 3 classes namely COVID-19, Pneumonia and Normal subjects.• Then 7 different CNN (Convolutional Neural Networks) models were also used, having a different number of convolution layers such as 2, 3, 4, 5, 6, 7 and 8 to build the same categorical classifier.• To gain more intuition about the dataset and understand the relation between COVID-19, Pneumonia and Normal patients, binary classifiers of each pair (COVID-19 vs Normal, Pneumonia vs Normal, Pneumonia vs COVID-19) were built using the same 11 pre-trained models and 7 CNN models as described above.• Obtained very high accuracies for COVID-19 vs Normal and Pneumonia vs Normal pairs, while for Pneumonia vs COVID-19 pair gained a little less accuracy as compared to the other two pairs. It shows that COVID-19 and Pneumonia patient’s X-rays have some features in common. • Next, to use the Machine Learning techniques such as SVM, XGBoost, Random Forest, LSTM, ANN, etc. more efficiently, instead of using direct Numpy arrays, features were extracted using pre-trained models (VGG16, InceptionV3, Xception, etc.) and then these Machine Learning techniques were applied on those features. Show less

• 

  TechJuly

  Aug 2020 - Sept 2020

  Machine Learning Intern

  • Built a predictive model which can detect Blur Images, using pre-trained models like VGG16, DenseNet121, ResNet, etc.• Built a Face Detection model using Face Recognition Library which extracts faces from the images and identifies the person in it.• Built a model using Facial Landmarks which can predict whether the Face in the image is tilted or not.

• 

  Turun yliopisto - University of Turku

  Jan 2021 - Mar 2021

  Deep Learning Intern

  Detection of Melanoma Skin Cancer disease through images using Deep Learning• Melanoma is the most severe form of skin cancer. It makes up 2% of skin cancers but is responsible for 75% of skin cancer deaths.• Distinguishing between malignant(infected) and benign(normal) cell is tough, as they both have almost same RGB color distribution.• Built binary classifiers using 11 pre-trained models(DenseNet, ResNet, etc.) and 7 basic CNN models and noted their accuracies.• Extracted features using pre-trained models and used ML techniques such as SVM, XGBoost, etc. for final classification. Show less

• 

  Udaan.com

  May 2022 - Jul 2022

  Data Science Intern

  Giving score to an Udaan Catalog Mario images according to certain quality parameters• Provides good user experience to the customers, helps in avoiding frauds on the platform, maintains uniformity on the platform, etc.• Quality parameters worked on - Image Resolution Check, White Background Detection, Center Alignment of the product Detection, 90 percent coverage of the frame by the product in either direction and Blur Image Detection• Used image processing techniques such as OpenCV python for detecting white background, center alignment and 90 percent coverage of the frame by the product• Built a classification model for blur image detection using CNN layers and pre-built architectures such as InceptionResNetV2 and InceptionV3• Identified priority verticals which should be targeted first for image scoring by calculating GMV value and number of impressions achieved by each vertical• Generated score of each image present in the vertical and drawn histograms and pie charts for visualization of scores• Tested the API of the project using Flask servers locally, deployed it using Docker and Kedro Framework Show less

• 

  Citi

  Jul 2023 - now

  Business Analytics Analyst

  Fraud Analytics and Strategy