ABOUT ME



"If you want to improve or challenge what is naturally available to us, the way is by going Artificial"

      — Puru Malhotra

Hi there! My name is Puru Malhotra. You have found yourself at the portfolio website of a passionate researcher, coder, and engineer!
I graduated with a Bachelor of Technology degree majoring in Information Technology in June 2021 with First Distinction (Highest Distinction) finishing in the Top 7% of the class.
I currently work at Deloitte USI (Tax) as a Business Technology Analyst. My past job experience was Artificial Intelligence Lead at Bizlers Technologies Pvt. Ltd.
My expertise covers software engineering in:
- Back-end (Infrastructure, Networking, Database)
- Front-end (Angular, Javascript, HTML/CSS)
- AI/ML.
I am fluent in C/C++, Python, C#, and Java(basic).
My recent work has been particularly dedicated to using technology to improve the lives of Persons with Disabilities and I am inclined to continue my efforts towards the cause.
I am always looking to work on exciting products and projects with others to learn and to improve as a developer and an engineer. If you have a position or idea that could use my experience, please reach out to me!

EXPERIENCE

Bizlers Technologies Pvt.Ltd.

Artificial Intelligence Lead

Was part of the core team of the Startup from the inception stage of their product TurboCare. TurboCare is a SaaS-based AI-powered automobile garage management tool for multi-brand garages available as both B2B and B2C applications. We have till date expanded across 4 cities across India and have served 3000+ customers along with 70+ partner garages. I managed all the Artificial Intelligence developments and worked along with attending full-time college.

  • Designed and coded pipelines to collect data, model architectures, and automation algorithms in python using TensorFlow, OpenCV and CNNs; Reduced manpower requirement from 3 people to 1
  • Developed flask REST APIs for number plate recognition, OCR, image segmentation (masking) for marking a region of interest in an image to identify the model of a car and priority sequencing to optimize the tasks in a job sheet associated with each automobile service.
  • Implemented text auto-filling based on learning using bi-directional LSTMs and bag of words from NLP for auto-generation of a service-related jobs sheet
  • Performed market research for 70+ customers to identify root level requirements. Prepared content for product brochure and investor pitch deck along with handling all the financial calculations. Participated in strategy R&D for developing a full-scale product
  • Led a team of 4 along with training 5 interns over the course of the duration of work.


    • Programming Languages: Python
    Technologies/Tools: AI (CV, CNN), ML(Random Forests, logistic             regression),Postman, Docker, Git, and Tableau.
    Frameworks: Flask, Tensorflow, Keras, OpenCV, NLTK
    Duration: May-September 2019 Artificial Intelligence Developer
         September 2019 - June 2021 Artificial Intelligence Lead



Deloitte Touche Tohmatsu Limited - USI Tax Technologies

Business Technology Analyst

Worked as a Developer in the National Federal Tax Services - Global investment and innovation incentives (Gi3) team.
Gi3 is an elaborate system of financial incentives to fuel companies’ R&D, innovation, sustainability and energy transition as well as other business investments exists in many countries and industry sectors.

  • Worked on research and development of an ML and NLP based keyword extractor. It automated the process of highlighting key phrases within an ESG document, a process which was done manually prior to this.
  • Worked on integrating archiveIT API into MyGi Deloitte web application to automate the document archival process for internal usage for 500+ clients in the US
  • Worked in Quality assurance to test 100+ web pages and documents. Further updated the web pages and documents tested to make them compliant to WCAG accessibility criteria using aria, lighthouse and WAVE tools to ensure accessibility.

Programming Languages: Python, C#, Javascript, Typescript
Technologies/Tools: SQL, Azure Cloud, Devops, HTML/CSS
Frameworks: .Net (Core and Framework), Angular, Tensorflow
Duration: July 2021 - Present

PUBLICATIONS

American Sign Language Identification Using Hand Trackpoint Analysis



Springer International Conference on Innovative Computing and Communications (ICICC) 2021

CONTRIBUTION: Yugam Bajaj and Puru Malhotra shared equal contributions as 1st Author

HIGHLIGHT: The study was done after formulating a problem statement based on personal interaction with 30+ hearing and speech impaired volunteers. It was aimed to design a two-way sign language translation system targeting the issue of sign language requiring a specialized interpreter, hence hindering day-to-day interactions. We designed an ASL recognition system working with an accuracy of 95.66%. We were able to achieve these results using simple Machine learning algorithms (kNN and random Forest) over complex CNN's generally used, making the technique lightweight in terms of processing power.

Virtual Meeting System Architecture with Gesture Control, Attentiveness Tracker and Automatic Attendance Recording

International Journal of Research in Engineering, Science and Management


CONTRIBUTION: Both Vatsal Gupta and Puru Malhotra shared equal contributions as 1st Authors

HIGHLIGHT: The system was designed aiming to improve the experience of virtual meetings which have become a new normal during the covid era for online learning. The system had 3 modules which included a customizable gesture-controlled module for functionalities such as mute/unmute mic, raise hand, etc. The Attentiveness tracker acted as both the warning system to the attendee and a check for the meeting host to ensure participants are attentive. The attendance marking system comes with a facial recognition system which ensures that only the designated person attends the meeting and marks his/her attendance automatically

Nano-robotic based Thrombolysis: Dissolving Blood Clots using Nanobots


2020 IEEE 17th India Council International Conference (INDICON)


CONTRIBUTION: 1st Author

HIGHLIGHT: The study proposes a medical procedure for the removal of blood clots. It discusses the design and working of Nano-robotic structures that carry medication directly to the affected area inside the body. The Nanobots find their advantage particularly in the cases where clots are located deep in the veins that are physically beyond the reach of doctors.

Linguistic Analysis of Tweets – Using Data Mining to study usage of English on Twitter


2020 IEEE 12th International Conference on Computational Intelligence and Communication Networks (CICN)

CONTRIBUTION: 1st Author

HIGHLIGHT: The study was conducted to understand the pattern of usage of English vocabulary on Twitter. It was inspired by an overly abbreviated text which took me multiple reads and internet searches to understand; which made me wonder how the use of English has changed. We concluded ~30% to be 'improper' and these improper words spread normally throughout the text. Hence, we contradicted some studies which suggested them to be occasional or accidental usage.

Mobility Assistive Device for the Visually Impaired




International Journal for Modern Trends in Science and Technology


CONTRIBUTION: 1st Author

HIGHLIGHT: The paper highlights the design of Mobility assistive glasses for the visually impaired. These are designed to assist locomotion in real-time. The glasses are equipped with sensors to realize the environment and guide the user through it by avoiding obstacles. They provide numerous advantages over the traditional walking sticks used as these are able to map the surroundings in 3D and are not limited by just range of reach

Nano Robots for Continuous Blood Glucose Diagnosis



International Journal of Trend in Scientific Research and Development (ijtsrd)


CONTRIBUTION: 1st Author

HIGHLIGHT: The paper discusses the continuous measurement of blood sugar levels through Nanorobots in the bloodstream. The bots are proposed to be having a Carbon-nanotube based structure. The method is less invasive, easier, and more reliable than the currently present methods

PROJECTS

Parselcode

  • Substitution based Encryption-Decryption Tool
  • Developed an algorithm by mapping word structures onto hindu birth chart houses
  • Package available on PyPi as Parcelcode
  • Technology: Python

DR. AI

  • AI based Medical app with Diagnosis and Prognosis feature.
  • Responsibilities: Project Lead, AI (datasets + models)
    Contribution- Built Flask API`s and Deep learning models for Chest X-Ray, MRI diagnosis (each having accuracy >82%) and ML-based Prognosis Models for Biliary Cirrhosis, Diabetic Retinopathy, Epidemiology and Lymphoma
  • App deployment halted subject to further testing due to ethical concerns to avoid spreding non-accredited medical results in public.
  • Technology: Python, AI (CNN), ML(SVM and Random Forest), Tensorflow, Keras, Flask, Docker

PDF Reader

  • Gesture controlled web based pdf reader.
  • Allows user to customise gestures for scroll up, scroll down, zoom in and zoom out
  • Technology: Javascript, HTML, Tensorflow.js

French Teaching Bot

  • Telegram chatbot with a backend powered using NLP and web scraping.
  • Makes use of Google API's to teach user pronunciations and vocabulary translations in french.
  • Data curated dynamically from web to enhance the topics for learning using a question answer based approach
  • Technology: Python, NLTK

Predictive Hindi Keyboard

  • Suggestive Hindi keyboard
    Using NLP to predict and giving user 5 possible choices as a next word based on what he has typed till now using a Hindi keyboard
  • RNN based architecture also using LSTM layers
  • Hindi translation of bible used as training dataset
  • Technology: Python, Tensorflow, Keras, NLTK

Lecture Notes Generator

  • System to record a live lecture and make notes by summarizing the content
  • Pointer generator based text summarizer involving bi-directional LSTM layers Speech to text module used to record lecture
  • An effort to help students concentrate on understanding the lecture rather than being busy in making notes
  • Technology: Python, Tensorflow,
Other Notable Projects: Image caption generator, Speech emotion recognizer, Estimating number of inversions, Recognize named entities on Twitter with LSTM, Self driving car controller, On-screen virtual keyboard.

CERTIFICATIONS


I never hold back on any learning opportunity and online learning has been a big support for me. I have completed 30+ certifications and specializations in a variety of topics from Coursera, Udemy, and LinkedIn Learning. The knowledge acquired was very helpful in bringing unconventional touch to my approaches as I gained a good root level understanding on the topics studied.

CONTACT