Dhairya Savaliya

About

Computer Engineering student at PDEU with hands-on experience in AI, machine learning, and data science. Built and fine-tuned models and projects across computer vision, NLP, and speech recognition using Python and modern ML frameworks.

Experience

• AI Engineer Intern
  TRYZENIQ
  Jan 2026 – May 2026
• Data Science & Machine Learning Intern
  BrainyBeam Info-Tech Pvt. Ltd.
  May 2025 – Jul 2025

Education

• Bachelor of Technology in Computer Engineering
  Pandit Deendayal Energy University (PDEU)
  Computer Engineering · 2022 – 2026
• Higher Secondary
  Mauni International School
  GSEB · 2020 – 2022

Skills

• Artificial intelligence
• Big Data Analytics
• C
• Deep Learning
• Express.js
• Git
• Google Colab
• Python
• Jupyter Notebook
• Machine Learning
• Linux
• MySQL
• VS Code
• Java
• JavaScript
• MongoDB
• Node.js
• NumPy
• Pandas
• scikit-learn
• OpenCV
• PyTorch
• TensorFlow
• PHP
• Docker
• Anaconda
• Hadoop
• Keras
• Postman
• React.js
• REST APIs
• Natural Language Processing
• Spark
• Software Engineering
• Kafka
• Hive
• Hugging Face Transformers
• C++
• Data Structures & Algorithms
• Database Management Systems

Projects

• Pathfinding Algorithm Visualizer
  Python, NetworkX, OSMnx, Streamlit, Matplotlib

  Designed an interactive visualizer comparing Dijkstra, A*, BFS, and Bellman-Ford algorithms on real-world maps. Integrated OpenStreetMap data with caching and dynamic graph visualization for algorithm performance analysis. Added route testing and performance comparison metrics, visualizing distance, execution time, and explored nodes.

• Deep Learning-Based Skin Cancer Detection with Grad-CAM (HAM-10000 Dataset)
  TensorFlow, Keras, OpenCV, Python

  Built a CNN-based classifier using DenseNet121 for skin lesion diagnosis on the HAM10000 dataset, achieving greater than 93% accuracy. Applied targeted data augmentation to address severe class imbalance across seven lesion types. Implemented Grad-CAM visualizations for explainability, validating the model’s focus on relevant lesion regions. Optimized training with stratified data splits and weighted loss functions for balanced performance.

• Fine-Tuned Speech Recognition System with OpenAI Whisper
  PyTorch, Transformers, Librosa, Gradio

  Fine-tuned OpenAI Whisper on LibriSpeech for English speech-to-text transcription, achieving approximately 90% accuracy (10.4% WER). Built an interactive Gradio web interface supporting GPU/CPU auto-selection for real-time inference. Implemented mixed-precision training, gradient accumulation, and a custom DataCollator for sequence padding efficiency.

• Sarcasm Detection in Gujarati Language
  Python, Hugging Face Transformers, IndicBERT, PyTorch

  Developed a sarcasm detection model for Gujarati text using the IndicBERT transformer, fine-tuned on regional-language datasets. Preprocessed data with tokenization, transliteration normalization, and text cleaning to handle Gujarati-specific linguistic variations. Evaluated model performance using precision, recall, and F1-score to ensure robust sarcasm classification in low-resource language contexts.

• Fairness Analysis of Face Recognition Models Under Environmental Degradations
  Python, PyTorch, ArcFace, AdaFace, OpenCV, Grad-CAM

  Conducted fairness evaluation of state-of-the-art face recognition models (ArcFace and AdaFace) across demographic subgroups under varying environmental degradations such as brightness changes and pose rotations. Analyzed recognition performance disparities using metrics including verification accuracy, false acceptance rate (FAR), and false rejection rate (FRR) to identify demographic bias patterns. Implemented Grad-CAM based visual explainability techniques to interpret model attention regions and investigate failure cases under challenging real-world conditions.

Courses & certifications

• Understanding Incubation and Entrepreneurship · NPTEL
• Python for Data Science · NPTEL (IIT Madras)
• Fundamentals of Machine Learning for Healthcare · Stanford Online (Coursera)
• Introduction to Deep Learning & Neural Networks with Keras · IBM (Coursera)
• Cybersecurity Fundamentals · IBM SkillBuild
• Artificial Intelligence / Machine Learning · Pregrad

🗣️ Languages

• english · Fluent
• hindi · Native
• gujarati · Native

🏆 Achievements & awards

• Ranked among the top 600 students in the entire state

📚 Publications

• Identification and Segmentation of Brain Tumors from Medical Images using Convolutional Neural Networks · 2025
• Deep Learning-Based Knee Injury Detection from MRI Images · 2026

🤝 Volunteering

• Lok Samarpan Raktdan Kendra · Community Service & Social Internship (CSSI) Volunteer · 2023

  Volunteered at a local blood bank as part of a college social internship initiative; assisted in donor coordination, data entry, and awareness activities promoting voluntary blood donation.