Machine Learning Engineer at Paris Beauty Cosmetics

Architected a multi-agent customer-care system using LangChain to route queries to specialized agents for recommendations (Pinecone with 500K+ embeddings), inventory (SQL), and orders (REST APIs), reducing response time from 15 minutes to 30 seconds. Built a production RAG pipeline with GPT-4, Claude, and Gemini APIs integrated with Pinecone/ChromaDB vector databases, powering intelligent beauty consultation chatbots with real-time <2s responses. Implemented an automated social media content pipeline using AutoGen’s GroupChat with a researcher agent (Reddit/TikTok APIs) and a writer agent powered by a fine-tuned LLaMA-2, producing 200+ brand-approved posts weekly with 3x engagement.

Embedded AI Intern at Kutta Technologies

Facilitated real-time object detection for UAVs in aerial search & rescue missions using Yolov8 on ARM Linux-based embedded Kutta SMC. Built Flask-based real-time visualization to display detection results (2s update interval) with OpenCV and PyTorch. Improved system accuracy and detection speed by ~50% by reducing false positives/negatives via CUDA optimization and faster frame processing. Devised multi-threaded asynchronous UDP video streaming with FFmpeg and OpenCV, achieving <50ms latency.

Machine Learning Assistant at Meteor Studio (ASU Research Lab)

Deployed computer vision pipelines and spatial AI algorithms using C# and Unity ML-Agents for immersive AR/VR applications. Conducted A/B testing and statistical analysis on user interaction data to optimize UX metrics. Transformed reinforcement learning systems for adaptive VR mechanics using deep neural networks and sensor fusion algorithms. Synthesized multi-modal ML models integrating visual, haptic, and spatial data streams. Deployed edge AI inference pipelines with optimized model quantization for real-time performance on mobile XR devices.

Machine Learning R&D Intern at Indian Space Research Organisation (ISRO)

Orchestrated a model for predicting downstream river stage over the Brahmaputra leveraging three parameters (water extent, precipitation, water level). Compared SARIMA and LSTM model predictions, achieving 97% accuracy. Prepared dataset for a Big Data project using geospatial satellite images (last 20 years). Augmented the dataset and performed Time Series Analysis with QGIS & Google Earth Engine, enhancing the prediction.

Data Engineer Intern at DocsVault

Developed a document capture module using ScanJob profiling tool to classify and extract information from scanned documents. Implemented it with .NET & MySQL, identified bugs and fixed 100+ issues. Used PySpark and CI/CD for large-scale data processing and analysis of document metadata, enabling better storage and retrieval.

AI Intern at SwissTech Unicorn Corp LLP

Collaborated on MERN stack development for the e-commerce app 'Astonpack'. Conducted Exploratory Data Analysis on transaction data using Python and Pandas to identify sales trends and customer behavior. Used Scikit-learn for predictive models to forecast product demand, improving inventory efficiency by 15%, and created interactive Tableau dashboards to visualize KPIs and sales metrics.

Software Engineer at Paris Beauty Cosmetics

Architected a distributed microservices platform using Kubernetes, Go, and React to power personalized beauty shopping experiences with AI-driven product recommendations, increasing average order value by 55% and supporting 2M+ daily active users. Designed real-time inventory optimization system with predictive analytics using Apache Kafka, Redis, and ML models that reduced stockouts by 65% and enabled dynamic pricing strategies to raise margins by 25% across 10,000+ SKUs.

Software Engineer Intern at Kutta Technologies

Facilitated real-time object detection for UAVs in aerial search & rescue missions using Yolov8 on ARM Linux based embedded systems. Built Flask-based real-time visualization to display detection results (2s update interval) with OpenCV and PyTorch. Improved system accuracy and detection speed by nearly 50% by reducing false positives/negatives through CUDA-optimized neural network inference and faster frame processing. Devised multi-threaded asynchronous UDP video streaming with FFmpeg and OpenCV, achieving <50ms latency.

Software Engineer Assistant at Meteor Studio (Verizon)

Utilized C# and Unity 3D to code app elements like physics simulations, player interactions, and particle behavior of multiple VR exhibits. Conducted iterative user testing and feedback sessions for refinement of game mechanics and interfaces. Implemented interactive VR mechanics for players to interact with virtual objects. Coded dynamic algorithms with adjustable parameters for user-customizable gameplay difficulty.

Machine Learning R&D Intern at Indian Space Research Organisation (ISRO)

Orchestrated a model for predicting downstream river stage over Brahmaputra leveraging water extent, precipitation, and water level. Compared SARIMA and LSTM model predictions, achieving an accuracy of 97%. Prepared dataset for this Big Data project using geospatial satellite images (last 20 years). Augmented the dataset and performed Time Series Analysis with QGIS & Google Earth Engine, enhancing the prediction and helping 5,000+ people by issuing flood alerts.

Software Engineer Intern at DocsVault

Developed a document capture module using Scanjob utility’s profiling tool to classify and extract information from scanned documents. Employed it with .NET & MySQL. Identified bugs and devised solutions to fix 100+ issues. Used CI/CD for large-scale data processing and analysis of document metadata, enabling better storage and retrieval.