Oct 2021 – Jan 2023 · 1 yr 3 mos · Ahmedabad, Gujarat, India

• Co-Created and Founded this growing Tech Community at LJ University with a mission to open opportunities for students to learn, collaborate and innovate on a common platform.
• Led a group of 15 people to handle the overall finance, marketing and functioning of the community along with its Social-Media and Online Presence.
• Created entrepreneurship opportunities for students by facilitating Innovation village where Startups and VCs were invited to gauge over 200+ ideas and concepts every year.
• Around 500+ students participated in events, workshops and bootcamps in this community, and this led to some amazing sponsorships for the community and internship opportunities for some talented individuals.
• My work involved curating and initiating sessions, arranging monetary funds, planning and streamlining strategies, hiring folks in the core team and negotiating partnerships, budgets and approvals for all the events.

Aug 2021 – Feb 2022 · 6 mos · Ahmedabad, Gujarat, India

• Worked upon 2 projects:
• INDEIGOGO
• 1. Developed a data modeling pipeline after scrapping 1 lakh projects from Indiegogo (A crowdfunding platform) to work upon determining creator credibility and success rates of projects.
• 2. For the first phase post cleaning and tuning the data, finalized 15 features that can reflect upon and define the creator's credibility. Gauged this with their previous successful projects and fundings.
• 3. Trained a semi-supervised Guided-LDA Model as the next phase that achieved 76% accuracy and almost 80% precision in predicting project-funding success based on the creator’s profile and their language (10 language/text features that can help determine the project's accountability) used for pitching the project.
• REDDIT
• 1. Scrapped 6000 reddits from the science handle to analyze user engagement across the platform for different topics.
• 2. Performed feature processing and kept 6 independent features (author’s fame, flair, and language features like centered-embeddings, capitalization in characters, use of jargons, numeric and special character values) in any post. Performed modeling with Decision Trees to predict user-engagement by gauging them with trust (used NC-Lexicons for calculating trust) developed from the user comments.
• 3. Got a 70% precision in likeliness of predicting engagement for reddits in the Science (Hard Sciences and Soft Sciences) Domain.

May 2021 – Jul 2021 · 2 mos

• Crafted a hybrid ML model that analyzed text descriptions leveraging attention mechanism to classify medical records into 10 disease classes.
• Got a classification accuracy of 75% when gauged with 2 medical practitioners for 50 reports.

Apr 2021 – Jul 2021 · 3 mos

• Deployed and got hands-on experience with Azure on Compute, App, Storage, and Security.
• Got trained on Windows Operating System, Active Directory, and networking concepts.
• Got a good understanding of business foundations and a good experience with approaching and solving technical issues for any business.
• Developed a project 'Koritsu' that is an online application to help boost team's productivity and pushing the team's morale while working at home with features like virtual attire, Microsoft groups, user-activity time reports, and a virtual check-in bot.

Oct 2020 – Dec 2020 · 2 mos · Ahmedabad, Gujarat, India

• Led the development of an automated camera assisted social distancing and mask detection software for a 2000-member workforce in a Chemical Factory.
• Built an intuitive UI utilizing JavaScript Stack that included the 2 major features and 8 camera streams with 15 FPS output for each stream.
• The ML models were built along with OpenCV's facial and body feature detectors to detect and analyze the key features with 95% precision.
• The project became a major contributor adding 20K$ to the yearly revenue.

Jul 2020 – Sep 2020 · 2 mos · Bengaluru, Karnataka, India

• Worked upon Diabetic Retinopathy grading and trained an image model with 10,000 images involving 5 classes
• Used CLAHE followed by median-filtering to preprocess the images that improved the accuracy by 70% to 85%
• Performed ROI cropping to reduce the image size by around 20% that reduced the training time by almost 30% followed by increasing the dataset size to 20,000 images to get a further 4% hike in the accuracy and reached a score of 89%

Feb 2020 – Apr 2020 · 2 mos · Ahmedabad, Gujarat, India

• 1. Developed a website prototype that helps users to get their own personal Global Identifier Number for their financial profile.
• 2. Performed market research and designed relevant content for the website's frontend to display features and benefits for the unique identification system.
• 3. Developed the database design and the APIs to connect the backend and frontend followed by hosting it at: www.globalidentificationsystem.com
• 4. Included features about shopping, purchasing or making payments with any of our partners and getting credits based on the purchases.