At Olympic Collectibles AI solutions (now Gradient), I focused on pushing the boundaries of Computer Vision and Large Language Model (VLM) applications to automate high-precision asset identification.

• Computer Vision & VLM Optimization: Engineered a multi-phase classification pipeline using DINOv2 (ViT-B/14) to identify card variants with 97.14% accuracy. I also researched and migrated metadata extraction to GLM-4.6V-Flash, achieving 98.6% accuracy for card number identification.

• Scalable AI Infrastructure: Developed a distributed inference system using FastAPI and Docker, utilizing round-robin scheduling to maximize GPU efficiency across worker nodes.

• Large-Scale Vector Search: Curated and indexed a 450+ GB dataset (150,000+ items) using Milvus and LakeFS, maintaining a 99.15% precision rate for “Best Fit” retrieval.

• Hardware-Software Integration: Solved complex environmental challenges like reflection detection by custom-configuring camera parameters (ISO, Shutter Speed) and implementing ambient light analysis.

In this leadership role, I took a generative AI concept from 0 to 1, focusing on rapid deployment and robust full-stack architecture.

• Platform Engineering: Spearheaded the design and launch of a generative video platform in just two months. I architected the full-stack system using FastAPI, Next.js, and Clerk for secure authentication.

• DevOps & Automation: Established a high-velocity development culture by building a CI/CD pipeline with GitHub Actions and Fly.io, significantly reducing manual deployment errors and streamlining release cycles.

At EcoHome, I am actively shaping user experiences and building technology-driven solutions to promote sustainability and community engagement. My work spans across key initiatives:

Alexa-Powered Inventory Automation

• Developed an Alexa Skill to extract receipts from emails and securely store inventory data with user consent.
• Engineered the backend using AWS Lambda, SQS, and MongoDB to ensure fast, reliable data processing.
• Integrated Gemini AI for intelligent data extraction and AWS Elastic Cache to optimize performance.

Non-Profit Donation Platform

• Built a React & Next.js web app enabling NGOs to request donated items from individuals and organizations.
• Designed a user-friendly interface to facilitate seamless donor-NGO interactions and streamline request fulfillment.

Key Technologies: Alexa Skill Kit, AWS Lambda, Amazon SQS, Gemini AI, MongoDB, AWS Elastic Cache, React.js, Next.js

Volunteered at IAN to support the development and optimization of the platform, progressively contributing as an Alpha Tester and Beta Tester.

• Identified and reported critical bugs affecting performance, security, and usability.
• Provided UX feedback to improve interface interactions (auto-scrolling, response delays, and navigation flows).
• Suggested and tested feature enhancements including query editing, search options, and admin notifications.
• Contributed to first-round documentation, detailing platform architecture, features, and usage guidelines.

Worked as Research Assistant at Sustainable Equity Allocation Resources SEAR Lab. Here I got exposed to ArcGIS a Geospatial Information Systems software and analytics platform. Improved my research skills and shared knowledge on GIS to peers.

• Conducted EV range analysis using ArcGIS Pro, mapping the maximum coverage area of various electric vehicles on a single charge across the U.S.
• Reconstructed 3D neighborhood layouts from 2D LiDAR data using deep learning techniques with 90% accuracy.
• Investigated and analyzed critical infrastructure in Texas, highlighting income and demography-based accessibility.

Key Technologies: ArcGIS Pro, GIS, LiDAR, Deep Learning

Worked in a fast paced early stage startup as a Scientist. Here I gained experience on agile development, developed knowledge in Computer Vision, researched different models, libraries to create prototype and deploy on the web.

• Deployed a computer vision model for post-joint surgery exercise analysis, serving 100+ patients and reducing additional physiotherapy costs by 5%.
• Achieved 95% accuracy interpreting medical device readings with AI, enabling real-time data transmission to healthcare providers.
• Built an NLP-based diet planner using 5,000+scraped Indian recipes, enhancing personalized nutrition plans.
• Worked with Python, Flask, AWS EC2, S3, and REST APIs to develop and maintain scalable software solutions.

Key Technologies: Python, Computer Vision, NLP, AWS EC2, S3, Flask, REST APIs

Interned at one of biggest FMCG’s and leading brand in the world. For the first time I exposed my computer science skills in data engineering and analytics in an industrial setting. I utilized the power of google suite to create a reporting system for anomalies in equipment in the factory.

• Created a web-based tool to streamline equipment maintenance reporting, boosting operational efficiency by 40%.
• Led training initiatives that improved workforce productivity by 15%, optimizing factory processes.