Default profile banner
PY

PARITOSH YADAV

@paritoshyadav

Data Scientist at Manifestit.io

Nagpur, Maharashtra

Manifestit.io (Sequoia Capital)

Paritosh Yadav is a Data Scientist and Machine Learning Engineer with extensive experience in Deep Learning, Computer Vision, and NLP. He has successfully developed and deployed AI solutions including LLaMa-2 ticket extraction, OCR systems, and recommendation engines. Paritosh has held key roles at Manifestit.io, AppZen, and ApparelTech, and holds a Master's degree in Computer Science and Engineering.

Experience

Data Scientist

Manifestit.io (Sequoia Capital)

•Nov 2001 - Present

LLaMa-2 Ticket Extraction(Application, Resource, Configuration). LLaMa-2 Effective Communication with resource configuration. Drift Detection in configuration.

Data Scientist

AppZen

•Apr 2021 - Aug 2023

Graph Attention Networks for Invoice Extraction. Custom NER Tagging (Natural language processing). Prompt Engineering and finetune on LLMA 2. OpenAI Model with Langchain (LLM). Transformer base model Key Value Extraction from any Invoices with 95% F1-Score. Serverless AI Function Deployment Using MLflow. Multilingual Modeling for Invoice(work for different languages). Document Element Extraction (table, text, title, image, Formula,List) and Apply Custom Rules.

Sr. Machine Learning Engineer

ApparelTech Pvt. Ltd

•Nov 2020 - Mar 2021

Smart-fit recommends best fit clothes to the customer base on their front side body measurements. Body Landmark Detection for calculation of body measurements trained model with adaptive Gaussian kernel, Deployed on AWS Sagemaker as Endpoint. Endpoint integrated in step function for for final output to device. Generate synthetic data using blender script and train model in Sagemaker to mask person (semantic segmentation) and linear Regression for body measurements.

Machine Learning Engineer

Trantor Software Pvt. Ltd.

•Jun 2019 - Nov 2020

Leveraged Multiple Instance Learning architecture for Spatio-Temporal Feature extraction in semi- supervised setting from videos and image sequences to identify processes from user activity logs. Setup detection and recognition model on sagemaker for training. Designed and Deployed an OCR solution in Sagemaker that outperformed the accuracy of Google’s Vision API by 7% benchmarked on ICDAR 2013 dataset. Face Recognition model finetune in sagemaker and deploy as endpoint for autoscaling.

Machine Learning Engineer

smartData Enterprises Pvt. Ltd

•Jan 2017 - May 2019

Handling large dataset for classification of which customer is going to make a claim in future based on their job description.(recall 0.91). Draw Generic machine learning architecture for classification problem. Based on Patent history, extract drugs, diagnosis and scan values, which help to get a clear and quick understanding of the patent to take further action. Deep Learning model for Damage car detection for micro insurance claim process. (detection+ localization).

Education

Master of Computer

Computer Science and Engineering

Jan 2015 - Jan 2017•Grade: 9.4 CGPA

M.C.M

Computer Science and Management

Jan 2014 - Jan 2015•Grade: 83%

Direct second Year

PGDCCA

Computer Science and Management

Jan 2013 - Jan 2014•Grade: 83%

Skills

Python
CPP
Embedded C
MLFlow
Deep Learning
Computer Vision
OpenCV
Graph Convolutional Neural Network (GCNN)
OpenAI
LLM
PyTorch
TensorFlow
CNN
Neural Network
NLTK
Scikit-Learn
Pandas
Numpy
AWS
GCP
Git
Docker
Lambda
Teradata
Step Function
MlOps
Sagemaker
Serverless Deployment
Smart Annotation