Harsh Patel
6261578341 # harsh21100@iiitnr.edu.in ï Linkedin Leetcode § Github
Experience
AI Engineer Intern Jan 2025 – Present
Meril life sciences
∗ Extracted patient data from pdf reports using PyMuPDF (text, 100%) and EasyOCR (image, 96%). Reduced
execution time from 37s to 16s, deployed on AWS SageMaker for scalable inference.
∗ Applied preprocessing and augmentation in CT Scan video dataset using Hough Circles, Hough Lines, Canny edge
detection, CLAHE, and other OpenCV functions boosting model performance by 10%.
∗ Developed an AI model for Aorta root bounding box prediction. Trained YOLOv8, RT-DETR, and a custom YOLO
model, achieving 0.89 mAP50 and GradCAM++ for explainable AI.
∗ Developed Aortic Valve Anatomy Prediction using video-based data for 3-class classification. Applied pre-trained
models (MC3, Efficient, Resnet,Two Stream ConvNet with optical flow) with best accuracy 88%.
.AI Engineer Intern Jan 2024 – Apr 2024
EVD technologies
∗ Built a Modern Football Chatbot using the Rasa framework, achieving an average response time under 1 second.
∗ Containerized application using Docker and deployed it on AWS EC2 with 99.99% uptime.
∗ Integrated real-time data from APIs and fbref.com into radar and bar graphs, cutting manual lookup by 80%.
∗ Implemented chat-based coaching session sched retrieval with 95% accuracy using PostgreSQL.
∗ Generated |3.8L in revenue for the company through US-based startup client.
Education
International Institute of Information Technology, Naya Raipur(IIIT-NR) 8.2/10
B.Tech in Computer Science Engineering Dec 2021 – Jun 2025
Academic World School 95%/100
CBSE - Class-XII May 2019 – May 2020
Publication
Stator Motor Fault Detection | Paper Link | Github |Autokeras, Streamlit, Shap, Lime, XAI Aug 2023 – Aug 2024
∗ Accepted at IEEE INDISCON 2024 PEC Chandigarh.
∗ Achieved 98% F1 score in detecting stator faults from vibration signals using an interactive UI pipeline with XAI tools.
∗ Created own motor vibration dataset consisting of 4 levels of severity of fault.
Projects
Multi Agent Customer Support Chatbot | Github | Langgraph, Gemini, MongoDB, RAG, PineconeMar 2025 – Apr 2025
∗ Built a website support chatbot for account setup, call scheduling, and other agents with <1s response time.
∗ Integrated RAG to deliver company info and recommend suitable medical products among 47 products of the company.
∗ Summarized chat history when message count exceeded 12, optimizing storage while maintaining context.
Fine-tuned Whisper ASR model for healthcare domain | Github | Transformers, Pytorch, Gradio Jul 2024 - Jul 2024
∗ Fine-tuned Whisper-small on domain-specific healthcare audio data, improving WER from 108.57 to 27.59
∗ Applied noise removal, speaker diarization, and resampling as preprocessing to improve transcription quality
∗ Implemented a real-time, speaker-attributed transcription interface using Gradio for seamless audio inference
Achievements
∗ Qualified GATE 2025 DA(Data Science and Artificial Intelligence) exam with rank of 2177 (top 4%).
∗ Selected for Amazon ML Summer School, earning a spot among 2,880 students from a 17,500 applicants.
∗ Solved 650+ DSA questions: Covering almost all the topics of DSA.
∗ 1,879+ rating(Knight) in Leetcode (top 5%.)
Technical Skills
Languages: Python, C++, SQL, Java, C.
Frameworks/Libraries: OpenCV, PyTorch, Transformers, BeautifulSoup, NumPy, Pandas, Scikit-learn, Flask, MLflow,
EasyOCR, MySQL, PostgreSQL, Langchain, Langgraph.
Developer Tools: Git, Docker, AWS EC2, AWS S3, AWS SageMaker, Ubuntu, Postman.
Positions and responsibilities
Shutterbug club: Photographed over 5 events at college, capturing key moments.
Technovate(Techno Cultural Fest) 2022: Organised various informal events having total participation of more than 500.
