Aera SDK• Engineered and integrated a Cortex AutoML client in Aera notebook, streamlining the ML workflow from data collection, model training to real-time inference.• Developed a flexible data retrieval solution with support for multiple storage clients for Aera Notebook, including SFTP and S3, incorporating robust security measures, authentication, and authorisation.• Drove SDK enhancements by allowing clients to define data metadata for existing CSV support and adding Gzip and Parquet file formats. This resulted in improved parsing efficiency for the Cortex system and simplified data cleaning processes for data scientists, facilitating smoother workflows for machine learning tasks.Aera SDK Operations• Created a robust health check framework for APIs, boosting operational resilience by ~60% and providing users with an intuitive interface for real-time diagnostics.• Designed an ETL pipeline using Django, to gather client usage metrics, creating graphical visualizations that facilitated strategic decision-making for management.• Engineered a robust SDK that independently reduced the ingestion time of 25M records to under a minute, enhancing memory efficiency and achieving a 80% boost in overall ingestion speed.

Intelligent-OCR System• Developed an ML system to extract information from documents such as Pan Card, ID cards, etc., and built a Named-Entity Recognition (NER) model using approaches like Bi-LSTM, RNN to parse the results, the same is used in the HR system.• Created a RESTful API-based web data extractor using selenium and used internally by SpringScan and deployed the solution using Docker.• Designed error codes and integrated error handling mechanism with Slack, thereby reducing 24 hours/month debugging time.• Stack: Python, NLP, Selenium, ECS, ECR, EC2, S3. Vision• Developed an ML model using Mask RCNN and GAN to extract land details based on an address using Google Maps API and achieved an approximate accuracy of 92%. • API performance and memory utilization were further improved by migrating the RESTful API from flask to fast API.Developed an algorithm to extract the geographical coordinates of the address and map them on the satellite imagery and vice-versa.• Stack: Python, OpenCV, pillow, Redis, Linux, ECS, ECR.Resume Parser• Developed the Resume Parser API using monolithic Architecture; however, efficiency has been increased and reduced response time by implementing multithreading and migrating into a microservice architecture. A reduction of 50% in response time was achieved.• Also, worked on react for creating a frontend website and hosting it on AWS.• Stack: Python, NLP, spacy, React, ECS, ECR, Cloudfront, AWS Lambda.

Resume Parser• Researched and worked on different algorithms to parse a document and analyze the word tags for further processing.• Advanced the accuracy and performance of the machine learning model and improved the accuracy from 65% to 82%.• Developed an end-to-end RESTful API to expose the above-developed functionality.• Stack: Python, GCP Compute Engine, NLP, Spacy.

• Worked upon various algorithms and approaches used to encrypt and decrypt files. • Developed a GUI by which a user can interact and do the decryption. • For decryption, a Fully Connected Neural Network was used and achieved a training accuracy of 90%.• Stack: Python, Google Colab, DSA.