
Hi, I am Prafful Varshney
Software Engineer
I am currently in my final year as an undergraduate student at IIT Roorkee,
pursuing a B.Tech degree in electrical engineering.
My academic and professional interests lie in problem-solving, software development,
machine learning, and deep learning. I have completed several projects in these areas, allowing me to apply
theoretical concepts to real-world challenges effectively. I am proud to have achieved notable distinctions in
competitive programming platforms. Specifically, I hold the title of Expert on Codeforces ,
5-star on CodeChef and Guardian on LeetCode .
Beyond my technical skills, I am passionate about continuous learning and growth. I am eager to take on new challenges
and contribute to the ever-evolving field of technology. Join me as I push boundaries and strive to make meaningful
contributions to the world of technology.
Experience
Software Engineering Intern @ Jio Platforms Ltd. May 2023 - July 2023
Contributed to CloudXP, a Hybrid-Cloud Management Platform, by automating the implementation of CIS benchmarks for MySQL databases. Enhanced MySQL database security by addressing data protection, compliance, and risk mitigation concerns.
- SQL
- Bash
- Shell-Scripting
- Python
Projects
Lip Sync Speech-to-Speech
Created a model to translate the audio of input video and sync the speaker's lips according to the translated audio using GAN-based networks to generate lifelike talking faces from the translated audio and input video.
- GANs
- Tensorflow
- CNN
- Deep Learning
Low light Image Restoration Challenge
Developed a generalized image enhancement model for images taken in low-light conditions. Used AHE and Gamma Correction to achieve an average PSNR of 26 on the LOL Test set. Won Gold medal for this problem statement in Inter-Bhawan competition.
- OpenCV
- Computer Vision
- CNN
Smart Horticulture Monitoring System
Developed an IoT‑based system to track the ambient conditions of the horticulture farm and automate horticulture process. Built a web app and integrated it with the ESP system using Firebase real‑time database.
- Arduino
- Firebase
- C/C++
- IoT
Stable Diffusion - Image to Prompt
Built a model to predict the text prompt for a given image using neural networks with NLP techniques. Implemented encoder-decoder model, featuring CNN network with ResNet50 weights as encoder for feature extraction and a standard LSTM network as decoder. Achieved a BELU score of 0.5.
- NLP
- Encoder-Decoder
- LSTM
- Deep Learning
Image Colorization
Developed an end-to-end deep learning pipeline to transform greyscale images into colorized images. Used conditional GAN conditioned on grayscale images, incorporated pre-trained ResNet-18 model as the backbone for downsampling in U-Net in Generator architecture and PatchGAN network as discriminator.
- GANs
- U-Net
- PatchGAN
- Deep Learning
Contact
Feel free to get in touch with me via email or ping me on social media.