Kiran Venkatesh

Experience

• 

  Indian Institute of Technology, Madras

  May 2018 - Jun 2018

  Summer Internship

  Worked in HTML,CSS,Tomcat ServerDone Courses in Machine Learning from Andrew NG and Dataquest

• 

  Kaleidofin Private Limited

  May 2019 - Nov 2019

  Research Intern-Data Science

  Anomaly Detection1) Identifies customer behaviour, data distribution for geography and agent behaviour pattern. Statistical Metrics are used to identify the anomalous portions of customers and agents in a specific branch. 2) Data Pre-processing,Cleaning,Visualization,metric generation & analysis are done to create a Quality Score. 3) Pandas, Seaborn are the tools used. Credit Analytics on Cashpor Data 1) Data Pre-processing,Cleaning,EDA & Survival Model are used for Cashpor data. Rest API is built for Survival Model with appropriate logger,timing functions & multiprocessing. 2) Univariate Cohort Analysis is the result of EDA & deployed the Survival Model as a microservices. 3) Tools used are Flask,GUnicorn,numba,pivot,binning. Done an excel model to evaluate responses and automated model using Python to calculate scores from responses for Financial Health calculator. Show less

• 

  IN-D - Power of AI

  Jun 2020 - Jul 2020

  AI/ML/CV Summer Intern

  Azure OCR & Image Analysis for Insurance 1) Image Preprocessing,Transforms & analysis are applied on Skewed,Rotated,Oblique images to convert different oriented images into proper images. 2)Image Preprocessing algorithms like rotation,skewing,homographic and perspective transforms are applied on the image before sending to azure ocr.These proper images are passed into Azure OCR. 3)Conversion of Google OCR to Azure OCR to work for any directory and trained for Receipts and Bills.4) OCR results gave extracted texts & corresponding Bounding Boxes. 5) Tools used are OpenCV,Pytesseract. Report Generation for Billspls 1) Conversion of JSON to an Excel Report & conversion of another JSON into an PDF report are done for Billspls details. 2) Tools used are Openpyxl, Reportlab. Show less

• 

  IN-D - Power of AI

  Jan 2021 - Jul 2021

  AI/ML Research Trainee

  Training VGG_16 classifier on KYC Documents 1) VGG_16 model loaded from existing hdfs file trained on old KYC Documents.Added checkpoints, image data augmentation like rotation range,zoom range,batch size to handle 90,270 degree rotated images for multi-class classification in existing VGG_16 code. 2) Boosted Accuracy from 93.5% to 99.9% to classify Aadhar,PAN,Driving License etc. 3)Tools used are Keras,Tensorflow,OpenCV. Field Classification of Extracted Text on Car Dekho Forms1) Image analysis on form samples & applied Google OCR to extract text. Extracted texts are classified into fields using various methods.Average Character Level accuracy is calculated to find which method suits for which field. 2) Achieved Average Character Level Accuracy of 50% to classify the extracted texts into respective fields. 3) Methods used are NER,RegEx & Fuzzy Token Set Ratio. Updated TensorFlow version to 2.x for mrcnn prediction module to do boundary detection on cards. Show less

• 

  WNS

  Jul 2021 - Jan 2023

  Analyst - Data Sciences

  Worked as a full stack data scientist at WNS FIAB(financial intelligence in a box) F&A analytics capability1)Exploratory data analysis(EDA),Data Wrangling & Pivoting of features are analysed in Account Receivables Collection Analytics using R dplyr,datatable.2)Algorithm implementation for WNS Predictor containing Summary Statistics,Graphical EDA, data visualization,Correlation plots ,feature engineering and selection, outlier removal using z score,handling multi-collinearity ,SMOTE,various ML models,model results & plots and inference predictions for new datasets.ML models implemented are linear regression,xgboost, Ann,svm, logistic regression, random forest, decision trees,adaboost etc. One hot encoding is done automatically in lm,svm package,label encoding for xgboost, Ann using data.matrix.Min max normalisation of data for ANN regressor.3)Data wrangling, combinations,jaro winkler,jaccard are used to calculate fuzzy similarity score for duplicate detection of large number of invoices. Included date and other columns using mapping in R.4) converted R code to python for duplicate detection fuzzy match using regex.5)Studied time series analysis and implemented stacked lstm in WNS forecaster. 6)Tools used are Plotly, Python & RSkills:- RShiny, GitHub, AWS, statistics, Time Series, data analytics, machine learning, NLP, Deep learning Show less

• 

  Ford Motor Company

  Jan 2023 - now

  Switched to operation analytics vehicle program health management team 0) Delivered MP2 timing slip regression based 2.0 models at upv0,upv2,TT, FEC,PP milestones using python & jupyter notebook 1) Handling backend of PRS LLM chatbot using GenAI and GCP 2) predicted mp2 slip days for upv0 milestone using alteryx, jupyter notebook and datarobot.ML models used:- gradient boosted greedy trees regression,feature selection,correlation analysis between target and feature,stepwise regression 3)Generated Predictions, Time Series plots and contours across various milestones for MP2 Timing Slip days. Automated the tasks using Python and Alteryx. 4)Downloaded of objectives, R/1000 & 3MIS CPU data from GSAR & decoder dump file.5) domain related training in vehicle program health managementSkills:- alteryx,data robot, python, jupyter notebook,VDI Show less

  • Operations Analytics Analyst - HTHD

    May 2023 - now

  • Operations Analytics Analyst - HTHD

    Jan 2023 - Apr 2023