About Me

avatar

Aakash Jhawar

A Data Scientist and a Full Stack Developer on a mission to joining the dots.

My Career

IDfy

Shared Services
  • Built custom authentication layer in user sign-in flow to validate the user with client’s token introspection service
  • Speed up the processes by automating workflows using inhouse CLI built on Thor gem in Ruby
  • Revamped Rate Limiting Service and boosted the efficiency to allow requests upto a threshold for individual clients
  • Maximized performance of API Platform microservices by redesigning the caching mechanism and flows
  • Conceptualized the end-to-end pipeline for storage on premise - built and deployed the services in the client’s on-premise environment
January 2020 - Present, Mumbai
Senior Software Engineer
Platform
  • Contributed to the concept of Multiverse (single and multi tenant) where the entire infrastructure can be deployed to a client’s environment
  • Developed infrastructure in a coded manner (Infrastructure as a Code) using Packer, Terraform and Kustomize
  • Implemented autoscaling of workloads based on queue depth and other metrics which reduced cost and manual intervention
  • Introduced Workload Identity and Cloud SQL Auth Proxy to help reduce the potential “blast radius” of a breach or compromise and management overhead
  • Carried out POC on Anthos by creating a cluster in AWS and managing the configurations from GCP
  • Shaped monitoring pipelines and alerts across the entire infrastructure using Likerd, Prometheus, VictoriaMetrics, and Grafana
January 2020 - Present, Mumbai
Senior Software Engineer
Machine Learning
  • Built Inhouse generic OCR pipeline using Mask RCNN, CRAFT text detection, Paddle text angle classification, Tesseract OCR and NER text classification
  • Improved Liveness model accuracy by from 78% to 97% using Transfer Learning
January 2020 - Present, Mumbai
Senior Software Engineer

IDfy

  • Built Aadhaar Number Masking Service to mask PII information on Aadhaar Card using Yolo V3 with an accuracy of 98.76% and a TAT of 1.04 sec
  • Implemented RabbitMQ message queues, created and maintained Docker containers on GCP clusters
  • Won the first prize in a 36hours internal Hackathon
May 2019 - July 2019, Mumbai
Machine Learning Intern

My Skills

My Projects

project.img

AvatarGAN

Generate Cartoon Images using Generative Adversarial Network. Generator learns to make fakes that look real. Discriminator learns to distinguish real from fake.

   ...  
project.img

Identify Commercial Centers using Point of Interest

Gathered all POI using Overpass Python wrapper. Preprocessed the dataset. Removed outliers from the dataset using DBSCAN. Created clusters based on amenities and plot the polygons on Google Maps.

   ...  
project.img

Face Recognition using OpenCV

Extracted face embeddings for each face in the dataset using pretrained OpenFace model. Trained a Neural Network to recognize faces with an accuracy of 90%.

   ...  
project.img

Fotoxo - Photo Management App

Implemented Sendgrid for sending account verification emails. Used Strip API for receiving card payments from users and AWS S3 Bucket for storing images in production.

   ...  
project.img

SolveSudoku - Sudoku Solver using DL and OpenCV

Extracted and segmented hand region from live video by thresholding. Count the number of fingers from the segmented hand region by using Convex Hull.

   ...  
project.img

Group Chat Web App

Achieved real-time messaging by implemented WebSockets with Action Cable. Used Devise gem for authentication. Used Semantic UI framework for front-end development.

   ...  
project.img

Investry - Stock Tracker App

A web app built using RoR in which users can track stock prices and follow particular stocks according to their choice. User can also connect other users and look up their stocks.

   ...  
project.img

Event Prediction using Sentiment Analysis

Developed a system that predicts the sentiment of Twitter users towards a specific event/keyword. Collected live tweets using Twitter Streaming API, used NLTK for tokenizing and TextBlob for classification.