Hi there, I'm

Daksh Loiya

Information Technology Student

Machine Learning & GenAI Enthusiast passionate about data analytics, AI, and cutting-edge technology.

About Me

I am an enthusiastic student at Manipal Institute of Technology, a premier technological institute in India, deeply passionate about data analytics, machine learning, and artificial intelligence I actively seek opportunities to apply and expand my knowledge through innovative projects and aim to contribute to cutting-edge developments in these transformative fields During my B. Tech in Information Technology, I built a strong foundation in software development, systems, and networking through courses like Data Structures, Database Systems, Java Programming, Operating Systems, Embedded Programming and Computer Architecture I specialized in web and app development and cybersecurity (including ethical hacking and AI in cybersecurity), data mining, and agile project management, and I took courses such as Thin Films & Vacuum Techniques.

Education

Manipal Institute of Technology, Manipal

Bachelor in Technology, Information Technology

08/2022 - 05/2026

During my B. Tech in Information Technology, I built a strong foundation in software development, systems, and networking through courses like Data Structures, Database Systems, Java Programming, Operating Systems, Embedded Programming and Computer Architecture.

I specialized in web and app development and cybersecurity (including ethical hacking and AI in cybersecurity), data mining, and agile project management, and I took courses such as Thin Films & Vacuum Techniques.

Work Experience

AI/ML Intern

ProsperaSoft

12/2024 - 02/2025 | Pune, Maharashtra, India

Developed Industry-Specific RAG Models: Spearheaded the creation of robust Retrieval-Augmented Generation pipelines using LangChain, FAISS, LLaMA, Transformers (BERT), Deepseek R1, and Groq to seamlessly integrate information retrieval and natural language generation—optimizing contextual accuracy, reducing inference time, and delivering precise, domain-tailored insights.

Trainee - Intel Unnati Industrial Training

Intel

05/2024 - 07/2024 | Remote

Experimented with various LLMs and fine-tuned Google's FLAN-T5 model on the DailyMail/CNN dataset using LoRA (Low-Rank Adaptation), a parameter-efficient fine-tuning method. Leveraging Hugging Face Transformers, Python, and advanced NLP techniques, developed and implemented a system that improved text understanding capabilities and generated concise, high-quality summaries.

Projects

Ongoing

RAG-Based Research Summarization using Monarch MONDO Ontology, with NER and Deepseek R1

01/2025 - Ongoing

Developed an Advanced RAG-Based Summarization Tool: Leveraged Retrieval-Augmented Generation (RAG) to automate concise summarization of biomedical literature, integrating the Monarch MONDO Ontology for semantic consistency and employing advanced Named Entity Recognition (NER) to accurately extract critical domain-specific entities.
Enhanced Precision with Deep Semantic Retrieval: Incorporated Deepseek R1 to perform deep semantic searches across large datasets, ensuring the retrieval of contextually relevant information and significantly reducing manual literature review time for actionable research insights.

Ongoing

Reinforcement Learning for Batch Reactors

01/2025 - Ongoing

Researched Advanced RL Techniques: Investigated the application of Deep Deterministic Policy Gradient (DDPG) and other reinforcement learning methods to enhance the accuracy of control policies, aiming to surpass the performance of the existing Nonlinear Model Predictive Control (NMPC) model.

Ongoing

Attribute-Based Access Control for Healthcare Data

01/2025 - Ongoing

Developed a Secure ABAC Framework: Designed and implemented an Attribute-Based Access Control (ABAC) system tailored for healthcare data, ensuring robust security measures and compliance with regulatory standards without revealing sensitive operational details.

Completed

Voice-Based Authentication System

09/2024 - 11/2024

Secure Voice Biometrics: Developed a secure authentication system utilizing SpeechBrain with automated one-time phrases and passwords, integrating AES and RSA encryption for enhanced data security.
Comprehensive Tech Stack: Built using MongoDB, Librosa, PyCryptoDome, Streamlit, Transformers, and Google Speech Recognizer to ensure scalability and user-friendly operation.

View on GitHub

Completed

Speech-Emotion Recognition System

07/2024 - 08/2024

Effective Emotion Detection: Achieved 60% accuracy on the CREMA-D dataset by leveraging LSTM models with MFCC for effective feature extraction.
Robust Data Processing & Visualization: Utilized Librosa, Pandas, and Numpy for data handling, along with Matplotlib and Seaborn for insightful visualizations.

View on Kaggle

Completed

News Article Summarization using FLAN-T5

05/2024 - 06/2024

Model Fine-Tuning: Fine-tuned the Google FLAN-T5 model on the DailyMail/CNN dataset using LoRA (a Parameter-Efficient Fine-Tuning method) during the Intel Unnati Industrial Training Program.
Advanced NLP Techniques: Leveraged Hugging Face Transformers, Python, and NLP strategies to improve text understanding and generate concise, high-quality summaries.

View on GitHub

Skills

Python Ecosystem

Python Pandas NumPy Librosa OpenCV

Front-End Development

HTML CSS JavaScript React.js Bootstrap

Back-End & Full Stack

Streamlit XAMPP PHP MySQL Node.js

Programming Languages

Java C C++ C# Bash OOP

Database Management

MongoDB SQL PL-SQL No-SQL

Natural Language Processing

FLAN-T5 BERT Meta LLaMA Deepseek R1 spaCy SciscpaCy NER NLTK

Deep Neural Networks

LSTM Optuna GridSearch CNNs RNNs ANNs YOLO

Reinforcement Learning

DDPG Q-Learning

Speech & Audio Processing

SpeechBrain Google Speech Recognizer Librosa

Information Security

AES RSA PyCryptoDome Hashing Encryption Biometrics Ethical Hacking

Certifications

IBM AI Engineer Professional Certification

IBM

13 Course Specialization in AI and ML

Inference on CPU and Fine-tuning of LLM Models to Create AI Custom Chatbot

Intel Corporation

HIPAA Research Module

CITI Training

Artificial Intelligence on Microsoft Azure

Microsoft

Deep Learning with Keras and Tensorflow

IBM

Introduction to Neural Networks and PyTorch

IBM

Generative AI Engineering and Fine-Tuning Transformers

IBM

Python Basics

University of Michigan

Extracurricular Activities

Editor-in-Chief and Head of Writing for The MIT Post

Served as Editor-in-Chief of The MIT Post, the official media organization for MIT Manipal, while holding an honorary position on the Student Council.
Directed content strategy and led the editorial team, enhancing the institute's media presence.
Coordinated communication initiatives that bolstered institutional branding and engagement.

Technical Head in Coding & AI/ML for IE-E&C, Manipal

Acted as Technical Head for the IE-E&C Manipal Chapter (IEI), driving key initiatives in coding and AI/ML domains.
Mentored junior members and coordinated project development, fostering a collaborative technical community.
Organized events and competitions that promoted innovation and skill development in emerging technologies.

Logistics & Supply Chain Management for Revels and TechTatva

Oversaw logistics for TechTatva and Revels at MIT, managing the supply chain for over 30 flagship events with tens of thousands of attendees.
Negotiated with vendors and administration to ensure efficient inventory and supplies management.
Streamlined operational processes to support large-scale event execution, enhancing overall event success.

Contact Me

Email

dakshhloiya@gmail.com

Phone

+91 9579009172

Location

Pune | Manipal, India

LinkedIn

linkedin.com/in/dakshhloiya

Hi there, I'm

Daksh Loiya

Information Technology Student

About Me

Education

Manipal Institute of Technology, Manipal

Bachelor in Technology, Information Technology

Work Experience

AI/ML Intern

ProsperaSoft

Trainee - Intel Unnati Industrial Training

Intel

Projects

RAG-Based Research Summarization using Monarch MONDO Ontology, with NER and Deepseek R1

Reinforcement Learning for Batch Reactors

Attribute-Based Access Control for Healthcare Data

Voice-Based Authentication System

Speech-Emotion Recognition System

News Article Summarization using FLAN-T5

Skills

Python Ecosystem

Front-End Development

Back-End & Full Stack

Programming Languages

Database Management

Natural Language Processing

Deep Neural Networks

Reinforcement Learning

Speech & Audio Processing

Information Security

Certifications

IBM AI Engineer Professional Certification

IBM

13 Course Specialization in AI and ML

Inference on CPU and Fine-tuning of LLM Models to Create AI Custom Chatbot

Intel Corporation

HIPAA Research Module

CITI Training

Artificial Intelligence on Microsoft Azure

Microsoft

Deep Learning with Keras and Tensorflow

IBM

Introduction to Neural Networks and PyTorch

IBM

Generative AI Engineering and Fine-Tuning Transformers

IBM

Python Basics

University of Michigan

Extracurricular Activities

Editor-in-Chief and Head of Writing for The MIT Post

Technical Head in Coding & AI/ML for IE-E&C, Manipal

Logistics & Supply Chain Management for Revels and TechTatva

Contact Me

Email

Phone

Location

LinkedIn

Daksh's Assistant