During my tenure at Perfios (previously Karza Technologies), I spearheaded the development of a cutting-edge model for text recognition in bank statements. By meticulously crafting and refining this model, I achieved a remarkable balance between low Character Error Rate (CER) and high accuracy in text box detection. This solution was seamlessly integrated into production, demonstrating its robustness and reliability.Additionally, I tackled the critical task of font classification and similarity analysis, particularly focusing on detecting tampering in KYC documents such as PAN Cards, Aadhaar, and Voter IDs. Leveraging advanced techniques including embeddings for cosine similarity checks, I consistently delivered high scores in top 1 and top 5 classifications. These accomplishments underscored my ability to navigate complex challenges in document verification with precision and efficacy.

During my tenure at Edubrite Systems, I played a pivotal role as an AI Engineer, focusing on innovative solutions to enhance learning experiences. One of my primary responsibilities involved developing smart question generation algorithms, leveraging knowledge graphs and other methodologies. These algorithms enabled dynamic generation of questions from diverse content sources such as paragraphs and courses, facilitating personalized and engaging assessments.Furthermore, I contributed to enhancing the platform's metadata suggestions for courses and lessons, optimizing the user experience through intelligent recommendations. By harnessing machine learning techniques, I empowered the platform to provide tailored recommendations, aligning with users' preferences and learning objectives. My contributions at Edubrite Systems underscored my commitment to leveraging AI for transformative educational solution

Currently working at Valuepitch E Technologies as a Data Scientist, working on problems on Computer Vision and Deep Learning such as face recognition on device i.e android using tensorflow lite and flatbuffers, working on image enhancement techniques for improving quality of OCR using tesseract and table extraction from pdfs, transliteration problems to facilitate conversion of pdf from local languages (such as hindi, marathi, etc) to english, in order to utilize NLP techniques for entity recognition, etc. Worked on deployment of APIs using flask for production on server. Have hands on experience with working with edge devices such as a raspberry pi for problems such as object detection. Working with deep learning based classification of audio data into different classes using mel-filter banks and spectograms and convolution based models.

Worked as a data scientist trainee on projects related to computer vision such as object detection, etc. Have worked on implementing person detection using Yolov3 on handlabeled images and also implemented frame by frame tracking of people using both CentroidTracker and Kalman Filter algorithms, have also implemented algorithms such as Faster-RCNN, opencv detection using HOG, dlib , etc. Also, created IoU metrics for performance monitoring of Yolov3 using keras callbacks.

I worked at Loyalytics Consulting ( a startup) in Bengaluru, India for 3 months as a Data Analyst intern from May 2018 till July 2018 end. As a data analyst intern, I worked on live projects from making dashboards for our clients. For making dashboards I employed tools such as SQL for importing the required data, R, python for data cleaning and wrangling and then Microsoft Power-BI desktop for making interpretable visualizations , so as to convey relevant information about the data.As a machine learning enthusiast I tried to create a standard churn model for prediction of whether a person has churned out of non-contractual services such as fashion retail business, electronics retail business, etc. Although, I wasn't able to create the model in the stipulated time, but learnt a lot in the process.