Karan Thacker

Hello and welcome! I'm a passionate technologist, dedicated to bridging the realms of Machine Learning, Data Engineering, and Software Development. My work experience and portfolio showcases an ensemble of projects, echoing my love for programming.

Work Experience

Apple, Inc

Cupertino, CA | 01/2022 - 08/2022 

  • • Built Machine Learning models by researching, implementing, and evaluating various published approaches for image processing.

    • Trained Deep Learning Model for Image Segmentation and object classification using PyTorch and using transfer learning from pre-trained Convolutional Neural Networks (CNN), thereby automating image editing tasks for Apple Media Products

    • Pruned and compressed Machine Learning models, improving prediction inference time by 50% and further reducing cloud computing resources.

    • Containerized (Docker) and deployed trained models on the cloud using AWS (EC2, ECR, Lambda, S3), creating a RESTful API, making the model available to be integrated as a micro-service

    • Gathered business requirements for automation/testing task at hand and carried out a preliminary analysis on best-suited Robotic Process Automation (RPA) tools available in the market for macOS and iOS devices

Infosys

Pune, India | 09/2017 - 01/2019 

  • • Contributed as software developer on an agile team, taking part in planning, analyzing, developing, testing and maintaining – Software Development Lifecycle (SDLC) of a product

    • Performed Data Migration and Remediation of Manufacturing Data in PostgreSQL, implemented QA automation and data analysis using R Programming, leading to creation of robust, optimized and fault tolerant data models for client

    • Created churn prediction POC on customer demographic data using ensemble and boosting decision trees in Python and Scikit-learn, giving 86% accuracy on test data, thereby helping client strategize targeted marketing campaign

Radiance Technologies

Ahmedabad | 05 /2020 - 08/2020 

  • Automated selection and ranking of bio-medical/clinical documents for medicinal research using Natural Language Processing models (spaCy, scispaCy) specializing in scientific text data, leading to 90%+ reduction of man hours spent on screening documents

    • Created data pre-processing and training pipeline on cloud ML infrastructure (AWS SageMaker), including tokenizing and vectorizing of text data

    • Deployed the ML inference pipeline on cloud, as containerized microservice application with Flask as the web framework

Indian Space Research Organization (ISRO)

Pune, India | 01/2022 - 08/2022 

  • Built a traffic load simulator to test satellite communication programs using Java and UDP protocols, to monitor performance and robustness of the system.

Skill Set

    • Python

    • MATLAB

    • R programming

    • C++

    • HTML

    • CSS

    • JavaScript

    • SQL

    • Docker

    • Kubernetes

    • Terraform

    • React.js

    • Node.js

    • Flask

    • Spark

    • MySQL

    • MongoDB

    • PostgreSQL

    • S3

    • EC2

    • Lambda

    • RDS

    • EKS

    • EMR

    • SageMaker

    • CloudFormation

    • Tableau

    • Power BI

    • Python (Matplotlib, Seaborn)

    • NumPy

    • Scikit-learn

    • TensorFlow

    • Keras

    • PyTorch

    • Pandas

    • Matplotlib

    • NLTK

    • Open-CV

    • PySpark

    • Jupyter

    • Notebooks

    • Git

    • Visual Studio

    • Code Google Colab

    • Apache Airflow

    • Hugging Face

Projects

Data Engineering and Analysis

  • Yelp Reviews Data Modeling

    100 Million+ records present in json format, parsed and cleaned using PySpark on Hadoop MapReduce cluster. Designed Normalised Data Model to store reviews and user information in Relational Database. Performed data analysis using SQL querying to gain meaningful and interesting insights.

    VIEW FULL PROJECT 👉

  • Twitter Scraping using Apache Airflow

    Created DAG process to pull latest tweets of specified User, organising in tabulated format and storing it in AWS S3 bucket using Apache Airflow and Python. The pipeline is hosted on EC2 and can be run as a CRON job on regular intervals.

    VIEW FULL PROJECT 👉

  • OpenStreetMap Data Analysis

    Parsed geolocation data of NYC present in XML format, fixed data inconsistencies, organised data from nested tags, storing in SQLite DataBase. Analysed data using complex SQL queries to get interesting trends and insights.

    VIEW FULL PROJECT 👉

Machine Learning

  • Image Captioning Project

    Trained CNN-LSTM cascaded neural network to generate text from image. Images were sampled from open source ImageNet Project. Model Architecture written in Pytorch and trained on GPU enabled cluster.

    VIEW FULL PROJECT 👉

  • Question Answering using Transformers

    Designed Encoder-Decoder attention network based on Transformer architecture from “Attention is all you need” research paper using Pytorch. Trained the model on open source HuggingFace Dataset.

    VIEW FULL PROJECT 👉

  • Image Segmentation

    Created AI Image tooling service to segment text out of image using CNN and inducing visual theme to segmented assets using GANs

    VIEW FULL PROJECT 👉

  • Enron Fraud Detection

    Applying Data cleaning , Feature Engineering and Machine Learning Algorithms on Financial Records and text obtained from mails to identify POIs in Enron Fraud Incident. The project was implemented using python and libraries including numpy, pandas and scikit-learn.

    VIEW FULL PROJECT 👉

Web Development

  • SmartFeed

    Web Interface providing personalised dashboard reflecting latest tweets from your selected topics and sentiment associated to it. The Application is hosted using Django Framework and AWS cloud services. Hosting and building of Application is automated using Terraform (Infrastructure as Code).

    VIEW FULL PROJECT 👉

Let’s Code Together