Jun 2019 – Oct 2021 · 2 yrs 4 mos

• Mainly working alongside the intersection of Computer vision and NLP.
• Built bank forms parser for cheques, mandates, etc. which is capable of extracting printed and handwritten information for each required field with more than 70% document level accuracy
• Built a generalized image denoising system which is able to remove noises or patterns in the image with minimal data.
• Achieved 80% document level accuracy on an in-house solution for unstructured legal documents which can identify name, address, case number, age, relationship for both accuser and accused
• Made customized solutions on real-time face recognition like intruder detection / Attendance system using person tracking, human detection along with FR.
• Worked on various preprocessing steps required for a Urdu OCR (Nashq and Nastaliq both) like horizontal & vertical separation of ligatures, dekerning, etc.

Jan 2019 – May 2019 · 4 mos · Gurgaon, India

• Worked on a product called VoC (recursive text clustering) used for getting marketing insights and resolved various issues including linked StackOverflow issue.
• Worked on code infrastructure development integrated with ELK stack for real-time modelling of data from Google analytics, google spreadsheets and google trends.

Sep 2018 – Oct 2018 · 1 mo · Noida Area, India

• Worked on an open-end research problem of building a CBIR for finding trademark similarity, and hence developed an ensemble model for ranking of a large pool of trademarks based on similarity to a particular trademark. Similarity detection was achieved by processing text, shape, color, objects, etc.

May 2018 – Aug 2018 · 3 mos · Noida, Uttar Pradesh, India

• Ad targeting: Youtube video classification into 12 genres based on title and description. Used a voting classifier on LSTM & SVM (as text is generally linearly separable) which takes Fasttext embeddings as input.

Nov 2017 – Feb 2018 · 3 mos · Bengaluru, Karnataka, India

• Involved in building a cognitive intelligent system in which my role is to make a machine learning based library for automation of the whole process from preprocessing of datasets to finally applying an algorithm to get results.

Sep 2017 – Oct 2017 · 1 mo · STEP, JSS-NOIDA

• Done a project based on prediction and analysis of breast cancer.

Jul 2017 – Sep 2017 · 2 mos · Noida, Uttar Pradesh, India

• Data collection and building REST API's