Higher Secondary

Bachelor of Technology - BTech

Part of the ML team to develop a virtual try-on system for clothes.
Performing literature survey and implementing state of the art algorithms on human body parsing, background removal, super-resolution among the various sub-tasks associated with building the same.

Machine Learning Engineer

Applied Scientist

Was part of the AI & ML team to extract entities from trade document data.
Used OpenCVfor pre-processing page images, used Tesseractfor OCR to convert the pdf into searchable pdf, and highlighted entities in the PDF using PyMuPDFlibrary.
Developed a lightweight UI using theStreamlitlibrary in Python for visualization of results

⋄ Contributed to the data science team in developing advanced deep learning models for tasks such as background removal and human body parsing, as part of a deep fashion project.
⋄ Finetuned ViT based Segformer model on semantic segmentation datasets achieveing IoU values of 58.02% on CIHP dataset, 64.70% on ATR dataset and 50.63% on LIP dataset.
⋄ Trained deep learning-based image matting networks specifically for human image datasets, optimizing background removal capabilities for deep fashion tasks.

Data Science Intern

Academically mentored 45 students, solved their doubts and conducted online tutorial sessions.
Conducted weekly live interactions and proctored quizzes and in-semester exams via Google Meet and the online platform SAFE.

MA106 Teaching Assistant

⋄ Developed trading strategies in the crypto market using Quadeye’s existing tech stack.
⋄ Implemented a hedging strategy between spot and futures of different underlyings to maximise profits.
⋄ Backtested the strategy on historic data by running simulations with appropriate constraints on gross, delta, volumes, positions, orders etc and ensuring liquidation of position upon expiry.

Quantitative Strategist

Worked on Excel Form Rendering, fetching dynamic form data using useHttp npm package.
 Used DataGrid component from MaterialUI for rendering the form and other custom MaterialUI input ele-
ments for the fields based on their types.
 Implemented custom validations to input fields of form by integrating the project with Formik and yup libraries.

React Developer

Worked on building personalised recommendations for the smart substitution model.

Data Analyst II

Was an L2 TA for the basic undergraduate CS course at IIT Bombay, responsible for mentoring the L1 TA s and filling in for their absences. 
Conducted project evaluations of around 15 students on MS Teams.

CS101 Teaching Assistant

⋄ Worked towards the development of a multi-lingual multi-modal question answering system for technical education courses offered on Spoken Tutorials, IIT Bombay, with the goal of enhancing learner engagement and knowledge retention through innovative technologies.
⋄ Constructed a weak supervision dataset by aligning questions from the Stack Overflow dataset with lecture transcripts from Spoken Tutorials, utilizing HuggingFace sentence transformers to extract relevant features for data alignment, facilitating effective training of the question answering system.
⋄ Conducted extensive literature survey and implemented diverse models, loss functions, and pre-training techniques to enhance model accuracy and achieve language agnosticism

Student Researcher

Applied Scientist II at Microsoft

Know about SOHAM MISTRI's work experience, education, bootcamps, certifications, and much more. SOHAM MISTRI