Python, C/C++, SQL, Artificial intelligence, Machine learning intern

Khushi Mahajan

+1-279-***-**** *************@****.*** linkedin.com/in/khushi github.com/khushi

Education

California State University Sacramento Sacramento, CA Master’s in Computer Science

Relevant Courses: Artificial Intelligence, Algorithms and Paradigms, Computational Biology Pursuing

Savitribai Phule Pune University Pune, MH, IND

Bachelor of Engineering in Computer Engineering

Relevant Courses: Machine learning, Deep learning, Data Science and analysis, Advanced Data Structures Aug. 2020 – May 2024

Experience

Data Science Intern Sep 2023 – Nov 2023

Atreya Innovations Private Limited (National Startup Awardee) Pune, MH, IND

• Optimized SQL and Python data pipelines for health monitoring systems, improving data retrieval speed by 35% and enabling faster analytics-driven patient insights.

• Enhanced data pipeline reliability, contributing to more accurate algorithmic health risk predictions used in wellness applications. Machine Learning Project Intern Jul 2023 – Mar 2024 Creocis Technologies Pune, MH, IND

• Engineered a predictive maintenance system by training an LSTM neural network using TensorFlow to forecast sensor values, achieving an RMSE of 0.069, and developed an SVM classifier (Scikit-learn) for real-time fault detection, attaining 96.8% accuracy.

• Implemented a scalable data preprocessing pipeline by analyzing a multi-rate time-series dataset consisting of 2,200+ instances, 43k+ sensor attributes, resampling inputs (1Hz–100Hz) into a unified 1Hz format and reducing feature preparation latency by 40%

• Developed a predictive maintenance workflow that forecasts faults 60 seconds ahead, allowing early corrective actions and improving operational efficiency and also designed a Power BI dashboard to visualize machine health.

• Deployed ML inference via Flask REST API and React dashboard, enabling dual manual/automatic modes for real-time fault detection and ML-driven corrective action suggestions, results published in Indian Journal of Technical Education, Publisher: Indian Society for Technical Education ISSN:0971-3034

Projects

AI-Powered Commentary System for Soccer Python, YOLOv9, OpenCV, NLP

• Engineered a multi-stage computer vision pipeline integrating YOLOv9, and fine-tuned a YOLOv9 model on 2,000+ annotated frames, achieving a mAP@0.5 of 0.901 for accurate player, referee, and ball detection in soccer broadcasts.

• Integrated ByteTrack algorithm into a multi-object tracking system achieving 92.7% accuracy in maintaining player identities across diverse sports scenes; decreased identity switches by 15%.

• Deployed a real-time pipeline capable of processing 23.6 FPS, synchronizing annotated video streams (bounding boxes, player IDs, possession stats) with generated audio commentary.

• Integrated computer vision outputs with template-based NLG and Google’s Text-to-Speech (gTTS) to generate synchronized audio commentary, enhancing accessibility for visually impaired users and enabling automated sports analytics. Intelligent Financial Advisor Chatbot Python, Retrieval-Augmented Generation (RAG), LLM, NLP, Qdrant Vector Database

• Architected a multi-LLM routing system using LangChain, dynamically selecting between Google Gemini Pro and Hugging Face model, improving response reliability and reducing fallback latency by 27%.

• Automated a RAG pipeline with Sentence Transformers and Qdrant vector DB, enabling secure QA over private financial PDFs and reducing retrieval errors by 34% compared to baseline keyword search.

• Developed and validated a sentiment classification model using XGBoost, performing feature engineering with TF-IDF on financial news text to accurately predict market sentiment, attaining 92% accuracy, and integrated it into a Streamlit full-stack chatbot with real-time data APIs. Sign Language Detection via Action Recognition Python, TensorFlow, Keras, PyTorch, Computer Vision

• Trained a deep learning algorithm using LSTM neural networks in TensorFlow and Keras for multi-frame gesture sequence classification, achieving 97% accuracy on a dataset of 1,000+ labeled gesture sequences.

• Optimized a real-time system integrating OpenCV for video capture and visualization, providing live webcam-based sign detection with responsive visual feedback to users.

Peer-to-Peer File Sharing System with Blockchain Logging Python, Flask, Solidity, MySQL, Docker

• Implemented a decentralized file sharing system supporting chunked transfers (up to 4KB/block), with peer discovery and a Flask-MySQL tracker successfully enabling real-time file lookup across nodes.

• Leveraged Truffle to deploy Solidity smart contracts for securely logging file events on a private Ethereum ledger and ensured data integrity via hash-matching and tamper-proof logging using Web3.py-integrated clients.

• Containerized and orchestrated 6 services using Docker, enabling full-stack deployment. Skills

Languages: Python, C/C++, SQL, MongoDB, Java, HTML/CSS Frameworks and Tools: React.js, Flask, Docker, Git, Google Cloud Platform, Power BI, Tableau, Web3.py, NumPy, CUDA, Pandas, TensorFlow, PyTorch, Keras, Scikit-learn, OpenCV, SciPy, REST, LangChain Interests: Object-Oriented Programming, Data Structures & Algorithms, Software Engineering, Data Science, Data Analysis, Natural Language Processing (NLP), LLM, Artificial Intelligence, Machine Learning, Deep Learning, Database Management System Involvements

• Delivered a seminar on “Handwriting Analysis for Detection of Personality Traits”

• Attended Honeywell Leadership Challenge Academy conducted at U.S. Space & Rocket Center, Alabama .

Contact this candidate