Pranjal Datta Yet Another Software Engineer

Resume

Contact Details

Work Experience

  • Machine Learning Engineer, Pixxel Space

    • Location: Bengaluru, India

    • June, 2022 - Present

  • Machine Learning Intern, Mad Street Den Inc

    • Location: Chennai, India

    • Jan, 2022 - June, 2022

    • Worked on optimizing and adding new features to the distributed workload orchestration platform.

  • Intern (Technical - Engineering), Synopsys Inc

    • Location: Mumbai, India.

    • May, 2021 - Jan, 2022

    • Designed novel ML-powered cloud-native tools to identify bottlenecks during chip manufacturing,processing over 100+ GB of data in realtime.

    • Rolling out the complete feature into production as an elegant, distributed software utilizing tools like Dask and Apache Spark with request-resolution time ~20% faster than legacy systems for large datasets.

    • Designing and building a pluggable Dask compute engine and a tool to manage Dask resources/Admin tasks/Scheduling for clean, extensible integration into the larger system.

    • Contributed to building the production CI/CD and deployment framework using tools like Jenkins, Kubernetes and Airflow.

  • Machine Learning Intern, OurEye.ai

    • Location: Bangalore, India

    • Oct, 2020 - Apr, 2021

    • Designed and implemented the production Machine Learning library from scratch.

    • Improved the end-to-end AI pipeline, resulting in ~30-40% higher system throughput and significant improvement in user experience leading to a ~25% increase in lead conversion.

Projects

  • PyVision: Built a toolbox of the latest computer vision algorithms, ready to use in 3 lines of code to reduce development time and allow for quick iteration.

  • Image2Sketch: Utilized Generative Adversarial Networks (GANs) to convert Selfies into artist sketches by training them on a custom dataset built specifically for the project. The video can be found here.

  • DenseDepth-PyTorch: Implemented the research paper to reproduce and verify the results of depth-estimation in 2D while modularizing the code following OOP principles for simpler usage.

  • AutoBlur: Designed a pipeline to to blur faces in videos automatically and improved it to handle the tricky edge cases. The key idea was to automate privacy protection in video clips.

Open-Source