Dec 2024 – Aug 2025 · 8 mos · Gurugram, Haryana, India · On-site

• Refactored the communication service to make it more extensible and integrated it with multiple vendors for SMS and WhatsApp messaging, reducing failures by 20%.
• Migrated user journeys from CleverTap to in-house backend by implementing notification workflows for order deliveries, reminders, and abandoned carts reducing journey costs by ∼60.3% while eliminating CleverTap dependency.
• Integrated the Firebase SDK to create a new push notification channel, resulting in a 10.5% increase in app ratings by targeting users who received orders but had not rated the app.
• Implemented email verification via OTP for sign-in/sign-up for the AI-Studio vertical.
• Co-ordinated with Front-End and QA teams to analyze and adapt features, successfully integrating solutions and resolving production defects collaboratively.

Dec 2023 – Dec 2024 · 1 yr · Noida, Uttar Pradesh, India · On-site

• Devised a data synchronization algorithm using timestamps to seamlessly sync data of 30+ contractor’s devices and the server, for distributed system data access.
• Refactored SQL queries for bulk operations optimizations, which decreased the device heating temperature nearly by 10◦ degrees.
• Built a end-to-end data collection framework for face-recognition models, automating processes like scraping, filtering, chip extraction, cleaning, and classifying.
• Conducted cleaning and classification using built framework for over ∼13 million face images utilizing custom filters and Amazon Rekognition, resulting in a final dataset of ∼1.5 million images across 5,000 identities with 100 images as cluster size.
• Developed business-specific, closed-domain face recognition model, trained on recognition images of employees, to achieve one-factor authentication with minimal false positives and ∼96.3% accuracy.
• Updated Face-Recognition Biometric App to dual-model based registration/recognition to further improve model accuracy from 96.3% to 99.84%.
• Updated in-field production device used for face registration/recognition with dual-model feature for over 25+ contractors.
• Engineered an automated system to continuously update facial registration encodings with clean recognition encodings, maintaining accuracy and preventing false negatives and false positives over time.
• Developed Aadhaar detection and OCR Service to get PII information from Aadhaar Card using Yolo V8 with an accuracy of 89%.

Sep 2023 – Dec 2023 · 3 mos · Gurugram, Haryana, India · On-site

• Implemented a chained task for automatic detection and deactivation of user vendor account, thereby
• revoking access to all vendor platforms upon user departure from the org after 3-5 days.
• Developed jobs for generating vendor reports, ensuring timely identification of all active, non-active, and non-registered users across multiple vendor platforms.