Jun 2021 – Aug 2024 · 3 yrs 2 mos · India

• Making it easy for Data scientists to build and deploy models at scale.
• Pieces I have been working on:
• Bring your own Image (BYOI), A tool that converts your existing code into a container, which you can test locally and perform and monitor training in a distributed environment.
• Provide the capability to Data scientists to write and test code locally, by simulating the prod environment, and one-click deployment.

Jul 2019 – Jun 2021 · 1 yr 11 mos

• ▶ Built a Production Information Manager system, which takes in transaction receipt and understand abbreviated and convoluted item names, ( Crml Lt -> Convert it into Human readable form, map it to Official Menu from Client, Categorise it to Coffee and Latte) . It leverages NLP Language Model (Trained on Food Understanding) and Knowledge DB and Databricks(spark) pipelines. Processes around 40Mn records each day. Received a Patent for the design and implementation.
• ▶ State of the Art NLP Classifier, Classifies a review into 7 classes which further leads to positive and negative in each class.
• ▶ Migrated previous machine learning models to Datalake and Databricks.

Nov 2017 – Jul 2019 · 1 yr 8 mos

• ▶ Goal was to achieve human level accuracy i.e around 85% and to manage a 3 member team. Successfully build a solution using Traditional Image Processing and Deep Learning simultaneously. We used object localization technique to track iris and optical flow to detect rotational movement. Result included torsional movement classifier with torsional time, angle, frequency and speed. It is in the testing phase at the client side.
• ▶ Goal was to achieve 95 % accuracy over polarity and auc score of 0.90 over 7 classes and their polarity. Supervised a 2 member team. Successfully build and deployed solution as API with inference average time 1 ms per review. We used Language Modelling approach and attention with biLSTM.

Jun 2016 – Nov 2017 · 1 yr 5 mos

• ▶ Building End to End Machine Learning Pipelines in NLP, Deep Learning and Predictive Analytics.
• ▶ Working on Object Segmentation with real time video feed. Working with Deep learning architecture like CNN, Faster-RCNN and making it production ready with TensorRT (Nvidia).
• ▶ Built a data extraction pipeline which extracts data hidden inside in tabular structure in Pdf forms / reports. Trained an CNN Model which classifies tabular structure. Further traditional image processing is applied to map data to it’s columns.
• ▶ Unstructured Documents(PDF etc) related to Health Sector(Clinical Trials) to extract both semantic information and structural information.
• ▶ Working on search engine specific to Clinical Trials and Clinical Reports. It uses semantic web technologies, Elasticsearch, Word Embeddings. Search Engine works on highly unstructured data (PDF, reports) etc.
• ▶ Languages worked with :- Python, Scala, Golang, Java, C++
• ▶ Tools which I have used :- ElasticSearch, Apache JENA, Apache-Spark, Hadoop
• ▶ Libraries and packages worked with :- Caffe, Keras, Tensorflow, Numpy, Pandas, Scikit-Learn, PdfBox, PdfMiner, Breeze, DeepLearning4j.

Feb 2016 – May 2016 · 3 mos

• Worked As Data Science Intern.
• Worked on Technologies like SPARK, Hadoop, H2o, Cassandra.
• Used Languages like Scala, GoLang, Python and R
• Worked on various use cases of ML for predictive and analysis use cases on Spark .

Oct 2017 – Oct 2020 · 3 yrs

Apr 2015 – Jul 2015 · 3 mos · Jaipur

• Worked as Backend deveoper (Intern) .
• Used GoLang, MongoDB and RehtinkDB.
• Implemented Spell Check and Suggestion feature for search .
• Implemented IP Throttling.