Tanisha Bhayani

Tanisha Bhayani roshantanisha.github.io linkedin.com/in/roshantanisha Education L. D. College of Engineering - Gujarat Technological University Bachelor of Engineering in Computer Engineering. CGPA: 8.22/10 Experience Freelance Machine Learning Engineer August, 2020-Present • I am working as Machine Learning Engineer for many different domains like health-care, real-estate, ML security, multi-modal models (CV + NLP), NLP, CV, Generative models, Online Learning, Outlier Detection. • Implementing research papers, providing consultation and what could be improved within the current research paper. Owing the complete end-to-end development pipeline in the ML projects. • Working on BERT, GPT based models, for solving tasks in NLP and NLU for not only english language. Optimizing already available code for increasing the performance of the model both in terms of accuracy and speed. Guiding developers on how to approach solving the client projects, solving their doubts, thereby reducing the development time by 80%. • Worked on property price prediction of canadian properties from the features of the property, demographics, socioeconomic, etc. Achieved an MAPE of less than 3%. Created graph based visualizations for better understanding of the results with respect to different pre-defined features. • Worked with companies like Intuit, and other niche startups for deploying their projects in the NLP domain. Also, worked with Schenck process to develop mobile/web application using React to collect annotated dataset, and then to train model in the domain of Computer Vision. Automated many tasks along the way. • Tutored over 100s of students to solve their doubts in Python and explaining complex concepts in AI, and related research papers. • Trained spaCy NER for financial document entity recognition, with 99% f1 score. • Created end-to-end AI web-application in Django that utilised langchain, openai model, for generating summaries from video. This involved working with several technologies that included in creating the pipeline for converting video to audio to text and then finally summaries. Along with this to keep the UX smooth, used celery, redis for background task processing, used google gmail API for email integrations. • Created real-time sign language detection model, used transformers model to classify the sign languages, and implemented innovative algorithm for doing real-time classification of the sign language from video stream. Visiting Faculty, Ganpat University August 2021-October 2021 • Teaching MSC-IT Students Machine Learning course. Teaching concepts like Clustering, Recommendation Systems, semi-supervised Learning, and ensemble methods. AI/ML Engineer, Silver Touch Technologies LTD January, 2020-August, 2020 • As an AI/ML Engineer, my job is to build an AI ecosystem in the company. • I develop end-to-end AI systems from requirements analysis, data gathering to deployment, implementing new methods/research papers, turning projects into research outcomes. Achieved substantial performance in DL/ML models for CV and NLP domain problems. • Hiring people for the AI engineer/intern role. Training and Overseeing their performance, and delegating tasks, and debugging any problem they face. Machine Learning Advisor, ArtuData (Through F(x) Data Labs Pvt. Ltd.) 2018-December 2019 September • Created different Machine Learning models for custom CRM data to predict the user conversion rate at different Tanisha Bhayani roshantanisha.github.io linkedin.com/in/roshantanisha steps of end-users. (Customer Funnel Prediction) • The model I created for a client was selected as best perfoming model among the top 3 companies competing. • Implemented an optimized search procedure for data augmentation for returning predictions and deployed it on GCP Server using Flask, Nginx, and VM. Associate Researcher, F(x) Data Labs Pvt. Ltd. June 2017-December 2019 • As a researcher, my job is to solve problems that are oriented with company’s goal. • I have achieved State of the Art results in problems like Facial Expression Detection and Recognition, optmization algorithms with best feature extraction techniques, and logical understanding of the data, and neural networks respectively. • Project planning, requirements gathering, analysing requirements to define architecture of the system, and timeline to implement them. Implementing and deploying projects, research papers as well as POCs. Creating presentations for the POCs to present it to the client. • Provide mentoring to junior developers for ML projects. Take daily standups, and provide suggestions for the problems. Do code reviews to improve the code-quality, and improvise logic via knowledge osmosis. Reverse Engineering Intern, F(x) Data Labs Pvt. Ltd. May, 2016 - August, 2016 • Worked on a custom database server project to create query graph for the buffer pool memory management for MariaDB. Implemented simple select, insert and delete query in MySQL 5.7.12 database server using C File Systems. • Understood and learnt MySQL DB software architecture from the source code. Helped and coordinated with my team members with understanding the code and concepts. Test Scores Test of English as Foreign Language (TOEFL) Score ETS.org November, 2020 Score: 106/120 Skills • Programming Methodologies: OOPS, Functional • Machine Learning and Data Visualization: tensorflow, PyTorch, scikit-learn, keras, NumPy, Pandas, seaborn, matplotlib • Machine Learning Deployment: Docker, Kubernetes, GCP, AWS • Databases and Languages: Python, PHP, JavaScript, MySQL, PostgreSQL, MongoDB • Software and Frameworks: Flask, Laravel, Angular, Jupyter, GitHub, git Research Work R Alogirthm Independent Research Work Tanisha Bhayani • I created a String Matching Algorithm that is better than Naive String Matching Algorithm by 40-50%, and 50% better than KMP algorithm in the worst case. Facial Age Estimation and Gender Detection Vikas Sheoran, Shreyansh Joshi, Tanisha Bhayani SilverTouch Technologies LTD Tanisha Bhayani roshantanisha.github.io linkedin.com/in/roshantanisha • Designed Deep Learning Model architecture for Age Estimation and Gender Detection. Ideated the use of relevant concepts of ML for the architecture and training process. Helped interns implement the architecture and taught them the concepts along the way. Reviewed the paper written, and approved the final paper draft as well. Achieved MAE of 4.67 on Age Estimation and Gender Detection accuracy of 95%. Paper got accepted in CVIP 2020 conference. Fake News Detection SilverTouch Technologies LTD Preksha Gandhi, Tanisha Bhayani • Detection of Fake News from News articles, getting 97% of precision score and 0.9922 AUROC score on 70k news dataset, where 32% were fake news dataset. Ideated the complete process and approach to solve the problem and helped the intern implement the solution comprising model architecture, data preprocessing with optimization to achieve the above value. F(x) Optimizer Algorithm F(x) Data Labs Pvt Ltd Hemen Ashodia, Tanisha Bhayani • Implemented and Tested F(x) Optimizer Algorithm, which is used for training Neural Networks. Created Reusable Python scripts for testing algorithm on various datasets, for comparing it with ADAM Optimizer. Selected Projects Generating creative writing for resume from short text Python, transformers, torch • Created a web-app (demo here: Demo) that would take short bullet point text that would then generate the resume experience that highlights skills and actions. • Fine-tuned DistilGPT2 model on resumes dataset. Conversion Funnel Prediction and Analysis Flask Python, scikit-learn, keras, matplotlib, Cython, • Created different Machine Learning models from custom CRM data to predict the conversion rate at different steps of end-users. • Won the best model (Customer Conversion Prediction) among top 3 companies competing. • Implemented an optimized search procedure (improved by 95%) for data augmentation for returning predictions, and deployed it on GCP Server using Flask, nginx, and VM. Loan Default Prediction (Research and Development) Angular, Laravel, Flask, ML libraries • Built a model to predict the propensity whether a loan applicant/holder would default or not ( 0.78 AUROC score). Created easy to use UI for filling details, and getting results in the form of graphs. Samyañc - My Spiritual Guru (Research and Development) Azure, Python, PHP, PostgreSQL NLP/NLU, Symbolic AI, Microsoft • ‘thelogicalkoan‘ is a closed domain question answering system in spirituality domain, based on the word-to-word English translation of Shrimad Bhagavad Gita. • The goal was to minimize the use of libraries, and research, design, and implement algorithms and system by own. System is tested with various questions to which it responds correctly. Created multiple unbiased data representations free from prejudice of the author by word-to-word translation of Sanskrit text in English for fetching the answer. Tanisha Bhayani roshantanisha.github.io linkedin.com/in/roshantanisha Medical Report Parser and Search (Development) NLP, Computer Vision, Python, PHP • Created a system to parse the medical reports in pdf, image and hybrid pdf formats. Also a web application was built to upload a pdf and get the parsed report in tabular form. Social Media Analytics (Development) Python, Cython, C++, Angular • Implemented Micro-Service Architecture, worked with different design patterns best suitable for this project. Implemented SOTA ML models with a tweak. • Developed DL model to parse the html tag that would contain text, with accuracy of 98% using custom collected data. • Mentored trainees for dividing further work and decided the workflow with tools and technology and steps at every point for different modules. Face Recognition System (Development) AngularJS Python, keras, matplotlib, scikit-learn, docker, • Created distributed design of the system, and integrated the SOTA ML models for face detection achieving 100% accuracy. Built ML model for Face Recognition with 96% accuracy on profile and frontal custom test dataset. Created an automation tool for data collection, training and a tool for manual incorrect data annotation. • Dockerized the project in terms of ML service and Front-end service which is in AngularJS. Mentored Projects • Facial Recognition for Banking System and automation • Fashion Trend recommendation • Stock Market Prediction using ANN • Fake News Detection (Research and Development) Ideation, Brainstorming, coding help Ideation, Brainstorming Research paper review MySQL, PostgreSQL, Python, Flask • Long Text classification for tender data Python, Jupyter, keras, BERT, sklearn, misc. libraries, Flask • E-commerce Product Recommendation System Python, pandas, pySpark, tensorly, numpy • Facial Age Estimation, Age Classification and Gender Detection Python, keras, pandas, numpy, sklearn, Flask • Study material recommendation based on student’s learning and cognitive patterns sklearn, xgboost, pandas, numpy, Flask Python, Awards and Mentions • Invited by Google Developers Group Gandhinagar to speak at DevFest 2022 on Federated Learning • Invited to speak at Goal Street International Workshop on Deep learning, shared the virtual stage with AI Chief Scientist of Reliance Jio. • Best student project award, 2017 for my final year project ’thelogicalkoan’ - Samyanc Mu Spiritual Guru. • Awarded with a book by the college for holding 1st rank in the department in the first year of BEng in 2013. • Awarded with prestgious award of Miss Xavierite for being consistent and outstanding student throughout schooling in 2012 • Was selected for the coveted Duke Talent Identification program on full scholarship to attend the Engineering class in 2009. • Awarded for winning the TATA Building India Essay Writing competition at City Level in 2008.