Green leaf




Harshit S - Python developer
Member since:Nov 22, 2019
Profile last updated:May 18, 2022
Last activity:Aug. 3, 2022, 2:26 p.m. UTC
Location:
India

Harshit S

I am a Machine learning and Data science, enthusiast. I like to solve problems related to computer vision, Natural Language Processing and Sequential models. I love working in a space where AI is creating great impact in helping businesses scale using automation.

Skills:    ·   ·   ·   ·   ·   ·   ·    ·    ·    ·    ·    ·    ·    ·    ·    ·    ·    ·    ·    · 
Weekly Availability: 56 hours
Day UTC PST
Sun 05 - 13 22 - 06
Mon 05 - 13 22 - 06
Tues 05 - 13 22 - 06
Wed 05 - 13 22 - 06
Thurs 05 - 13 22 - 06
Fri 05 - 13 22 - 06
Sat 05 - 13 22 - 06

Hourly Rate: $27.50
Experience: 3+ yrs
Python: 3+ yrs
Machine learning: 3 yrs
GPU accelerators: 2 yrs
C++: 2 yrs
Computer vision: 1+ yrs
Flask: 1+ yrs
Natural language processing: 1+ yrs
Engineer's Devices:
Mac
Android

LD Talent Work History

Average response time:
4.1 hours
Avg count of messages / Day:
6.1
Average client satisfaction:

Earned hours:
478.50
Worksession approval:
100.0%
# Hires / # Interviews :
6 / 10
# Passion projects:
1
 

Experiences
UW-Madison-CMS     UWMadisonCMS     Developer     Contract
Dec 2020 - Jun 2022
Research

  • Developing a Machine Learning algorithm with a viable FPGA implementation for detecting novel particles with data streaming in the L1 Calorimeter Trigger region, CMS at a rate of 25ns/event.
  • Handled model design, training, and simplification for FPGA implementation. Currently, we are researching Autoencoder (AE)/ Variation autoencoder models with semi-connected networks for detection.
  • Also worked on Simulation of Various High energy Physical Processes using C++ and Conducted GPU runs using CUDA codes over GPU accelerators

Skills used: GPU accelerators, C++, Machine learning


WriteWise     Developer     Contract
Oct 2021 - Jun 2022
Education

  • I am responsible for python development in Flask and Handling of the Artificial Intelligence backend
  • Built propriety features using Natural Language Processing concepts for WriteWise API
  • Management and development of pretrained models using the Tensorflow (v1) framework

Skills used: Python, Flask, Artificial intelligence, Tensorflow, Natural language processing


Ingage     Developer     Contract
Dec 2021 - Apr 2022
Technology

  • Implemented state of the art smiles detection deep learning solution and fit the model to specific company data
  • Worked on developing emotion detection algorithm using machine learning approaches over audio data
  • Researched and Implemented Computer Vision Methods for speed decode for DL inference time optimization

Skills used: Computer vision, Deep learning, Machine learning


Lattice Data     Developing next-gen Financial Data API Services     Developer     Contract
Jan 2021 - Jan 2022
Finance

  • Developing smart API components using NLP techniques over Python Flask to process and analyse open text data such as stocks-related news and forums discussions.
  • The lead manager of product development pipeline and server over Google Cloud Platform. Implemented Continuous Integration and Git contribution strategy.

Skills used: Google Cloud, Flask, Python



Link to the project:
Learning Dollars     Developer     Contract
Jul 2021 - Dec 2021
Finance

  • Working on developing Machine Learning models for stock growth prediction
  • Used scikit-learn to develop basic models for prediction over supervised format data
  • Used concepts of Natural language processing (Sequential models) to model time series data

Skills used: Machine learning, scikit-learn, Natural language processing


Indian Institute Of Technology, Jodhpur     AI Driven Estimation of COVID-19 Prognosis and Comorbidities using Multimodal Data     Student     Internship
Nov 2020 - Nov 2021
Healthcare

  • Improved model performance from 0.79 AUC score to 0.85 AUC score over test dataset for severity level prediction and obtained a prognosis score of 0.88 over test dataset (BIMCV COVID-19) dataset.
  • Generated and trained AE and VAE Computer Vision models based on Resnet, UNet, and Densenet to encode information to lower dimensions for feeding an LSTM based prognosis model.
  • Designed and deployed custom online triplet loss, implemented perceptual loss using pre-trained VGG network and custom weighted loss using MSE, Perceptual, and SSIM loss.
  • Custom coded the architecture, losses and preprocessing stages using Keras Framework in python.
  • The project was sponsored by DST India and was published at the IEEE ISBI 2022 conference.

Skills used: keras, Computer vision, Python



Link to the project:
Docket Science     Self Drive Route Optimization     Developer     Contract
Apr 2020 - Jul 2020
Research

  • Worked On Big Data to optimize path routing by appropriate algorithm development and Multithreaded Programming on Python
  • Performed data analytics to help develop algorithm and routing strategy
  • Worked with pandas as data handling module for algorithm developed and for data analytics
  • Worked on producing dynamic routing graph animations using Matplotlib for better result feedback.

Skills used: Python, Multithreaded Programming, Pandas, Matplotlib, Data Analytics


Blossom     Developer     Contract
Jan 2020 - May 2020
Technology

  • Worked on company obtained data to produce a machine learning model to predict company attributes
  • The Project Required extensive use of Natural Language Processing to complete a Multi-label operation
  • Python was employed to write the data processing, training and prediction scripts
  • Worked on company data on MongoDB to Pull and push results to the development pipeline

Skills used: Natural language processing, MongoDB, Python, Machine learning


Indian Institute Of Technology, Jodhpur     Credit fraud detection using sparse Gaussian processes     Student     Internship
Sep 2019 - Jan 2020
Finance

  • Used the concept of SVGP approximation (Bayesian Machine Learning) to detect credit fraud activity and develop a technique to work with an imbalanced dataset to provide optimum inducing points.
  • Hyperparameter optimization and varying inducing points yielded the best results using the RBF kernel with 0.98 per cent accuracy over the test dataset.
  • Developed experiment using gpflow and scikit-learn in Python

Skills used: scikit-learn, Python, Machine learning



Link to the project:
Learning Dollars     How to create a language translator using Tensorflow.JS     Developer     Passion Project
Dec 2019 - Jan 2020
Education

  • Coded the training procedure of the language translator in python
  • Deployed the machine learning training procedure using TensorFlow Framework
  • Converted the model to TensorFlow-JS architecture, imported weights and deployed over Nodejs

Skills used: Nodejs, Tensorflow, Python, Machine learning



Link to the project:
Indian Institute Of Technology, Hyderabad     Alien Object Segmentation Using Deep Bayesian Methods     Student     Internship
May 2019 - Jul 2019
Research

  • This project involves the use of Deep Gaussian process to perform segmentation by also marking the uncertainty of the prediction
  • The tensorFlow framework was used to perform the experiments and train the Machine Learning Models
  • The experiments were performed over NVIDIA-DGX GPU accelerators server of the institute

Skills used: Tensorflow, GPU accelerators, Machine learning


CoStacks     Crops yield estimation and selection : Web Application     Developer     Contract
Apr 2019 - Jun 2019
Food and Agriculture

  • The complete project from training to web deployment was done using python
  • The data was refined and additional data for researched for to improve the input to our models. Data analytics were derived from the data to form meaningful features for the model input
  • The training procedures were deployed using scikit-learn functions.
  • The training was performed using machine learning algorithms such as decision trees, random forest, etc, choosing the best models for ensemble.
  • The web application was presented to the client deployed using Django

Skills used: Python, Django, scikit-learn, Data Analytics, Machine learning



Education
Indian Institute of Technology, Jodhpur    B. Tech  (Electrical Engineering)
Jul 2017 - Jul 2021

Courses
B.Tech Project : Sem II    IIT Jodhpur
Jan 2019 - Apr 2019
Skills learnt: keras, Deep learning, Machine learning
Digital Image Processing    IIT Jodhpur
Jan 2020 - May 2020
Skills learnt: Computer vision
Statistical Signal Processing    IIT Jodhpur
Sep 2020 - Nov 2020
Skills learnt: scikit-learn, Data Science, Machine learning

Client Reviews
Ingage:    
Harshit exhibits an extraordinary amount of engineering judgement overlaid with his subject matter expertise in machine learning. Wildly impressed by his competence, communication and clarity of thought. Most developers just execute, but Harshit can only do that with immense precision but also be a thought partner. Loved working with him and look forward to doing so again.
WriteWise:    
Harshit is an amazing engineer. He is proactive in everything he does and carefully detailed in his work. His communication with the management team is clear and prompt, ensuring that project goals are met to or exceeding expectations. I will be happy to continue working with Harshit on all of our ongoing projects!
SearchCo:    
Nice work
LD Talent:    
very impressive work - you solved the issue I had, and also made the code go faster as well
UWMadisonCMS:    
Does a good job. Knows a thing or two about GPUs.
Blossom:    
Faster execution!