Contact

About Me

I am an AI & Machine Learning engineering student with hands-on experience in building end-to-end machine learning and NLP systems. I specialize in designing data pipelines, fine-tuning deep learning models, and applying modern NLP and large language models (LLMs) to real-world problems.

My interests include Natural Language Processing, Large Language Models, federated learning, and applied AI systems. I enjoy experimenting with models, evaluating performance, and building scalable and practical solutions.

---

Experience

Research Assistant Intern – Woxsen Agentic Lab

Dec 2025 – Present

  • Fine-tuned open-source large language models for agent-based AI tasks using LoRA and QLoRA
  • Conducted multiple controlled experiments and analyzed model performance to support optimization decisions
  • Collaborated with a research team to improve LLM training and evaluation workflows
---

Projects

🔹 Multimodal Image Captioning

GitHub: https://github.com/sathvik-web/multimodal_image_captioning

  • Developed a multimodal deep learning model that generates descriptive captions for images by combining computer vision and natural language processing techniques.
  • Extracted visual features using a pre-trained CNN encoder and generated captions with an LSTM-based sequence decoder.
  • Built an end-to-end pipeline including image preprocessing, feature extraction, tokenization, and sequence generation.
  • Evaluated caption quality using BLEU scores and qualitative analysis to assess language fluency and relevance.
---

🔹 Handwritten OCR System

GitHub: https://github.com/sathvik-web/Handwritten-OCR-System

  • Built a handwritten Optical Character Recognition (OCR) system to recognize handwritten text using deep learning techniques.
  • Implemented image preprocessing steps such as noise reduction, normalization, and segmentation to improve recognition accuracy.
  • Trained a convolutional neural network (CNN) model to classify handwritten characters from image inputs.
  • Developed an inference pipeline that converts handwritten images into machine-readable text for automated document processing.
---

🔹 Voice Authentication Anti-Spoofing System

GitHub: https://github.com/sathvik-web/voice-authentication-anti-spoofing-system

  • Developed a voice authentication system capable of detecting spoofed or synthesized audio used in voice-based security systems.
  • Extracted audio features such as MFCCs and spectral representations for training machine learning models.
  • Built a classification model to distinguish between genuine and spoofed voice samples.
  • Designed the system to enhance security in voice biometric authentication applications.
---

🔹 RAG-Based PDF Question Answering Bot

GitHub: https://github.com/sathvik-web/rag-pdf-qa-bot

  • Built an intelligent Retrieval-Augmented Generation (RAG) system to answer user questions from PDF documents
  • Implemented document embedding, vector search, and context retrieval for accurate response generation
  • Designed a modular inference pipeline suitable for real-world document analysis and research workflows

Key skills: NLP, embeddings, vector search, LLM inference, pipeline design

---

🔹 Abstractive Text Summarization using T5 Transformer

GitHub: https://github.com/sathvik-web/text-summarizer-app

  • Fine-tuned a pre-trained T5 Transformer model to generate human-like abstractive summaries
  • Built a complete NLP pipeline including text cleaning, tokenization, attention masks, and sequence padding
  • Evaluated summarization quality using ROUGE metrics and manual readability checks
  • Developed a simple Python interface for model inference

Key skills: Transformers, NLP pipelines, evaluation, deep learning

---

🔹 Federated Learning using CNN

GitHub: https://github.com/sathvik-web/federated-learning-chestxray

  • Built a federated learning system in which multiple clients collaboratively trained a CNN model without sharing raw data
  • Designed client-side training workflows and simulated multiple federated learning rounds
  • Implemented FedAvg aggregation and analyzed convergence behavior across training rounds

Key skills: Distributed learning, CNNs, privacy-preserving ML, PyTorch, TensorFlow

---

Skills

Programming: Python, Java, JavaScript

Machine Learning & AI: Scikit-learn, TensorFlow, Keras, Feature engineering, Pytorch, Pandas

Cloud & Databases: AWS, Docker, POSTGRESQL,

Tools: Git, GitHub, Jupyter Notebook

Other: Problem solving

---

Patent

Autonomous Flood Monitoring System (Design Patent – India)

  • Co-invented and registered an autonomous flood monitoring system
  • Designed a solution integrating environmental sensors, solar power, and wireless communication
  • Developed for real-time flood monitoring and early-warning support for disaster management
---

Education

B.Tech – Artificial Intelligence & Machine Learning
Woxsen University, Hyderabad
GPA: 8.61 / 10
Duration: 2023 – 2027 (Pursuing)