Hello! I'm Ayush Saha, a Computer Science and Engineering student at VIT-AP University.
I
With strong interests in research, Generative AI, web development, and art, I enjoy exploring both technology and creative expression.
Skills
Education
- Computer Science and Engineering | VIT-AP University (Sep’23 – Present) | CGPA: 8.64
- XII (CBSE) | Sri Sri Ravishankar Vidya Mandir | 87.8% (2022)
- Art School | J.B. School of Fine Arts (2008–2020)
Experience / Research
Explored human settlement on Mars, focusing on astrobiology, in-situ resource utilization, environmental challenges, and habitat development. Evaluated current technological approaches and sustainability strategies for long-term colonization. Contributed original analysis and conceptual frameworks. Co-authoring the paper “Mars: Exploring Life on the Red Planet” (under review).
Worked on Generative AI, PyTorch, Scikit-learn, Java, and R. Building supervised and unsupervised ML pipelines, fine-tuning models, and integrating AI solutions into automation workflows. Selected based on skills and two rounds of interviews.
LinkedIn View Certificate
Contributing to AI/ML initiatives within the GDG VIT-AP community through data analysis, dataset preprocessing, and machine learning experimentation. Working with team members on AI-driven projects and exploring practical applications of machine learning and data analytics.
View Certificate
Certifications
-
IBMCE_CEWXAI1IN Certificate
Completed IBM’s “Generative AI Using Watsonx (CEWXAI1IN)” course. Gained skills in prompt engineering, fine-tuning, and deploying GenAI models using Watsonx.ai & Watsonx.data with a focus on responsible AI.
View Certificate -
Wadhwani Ignite 3.4 – Entrepreneurship & Innovation Program
Completed a 14-week program focused on entrepreneurship, startup ideation, MVP building, and investor-ready pitch decks.
View Certificate -
AWS Academy Cloud Foundations Training
Gained foundational knowledge in AWS services (EC2, S3, IAM, VPC), cloud architecture principles, and security best practices.
View Badge -
AWS Academy Graduate - Cloud Architecting Training
Hands-on experience with AWS cloud services, scalable architecture, and secure cloud solutions.
View Badge -
IndiaAI Impact Gen-AI Hackathon – IISc CNI & IBM Research
Shortlisted nationwide for a 4-week Generative AI program. Gained hands-on experience with datasets, open-source tools, IBM ADK, WatsonX, Kaggle, and GitHub Classroom.
View Post
Academic Projects
Women Safety Backpack | VIT-AP University (Nov’24)
Developed a smart safety backpack using ESP32, GPS, and a mobile app for real-time alerts and tracking. Emergency buzzer and Bluetooth connectivity included.
Technologies: ESP32, GPS, Bluetooth, Mobile App Development.
Why I built this
I built this project to explore how technology can be used to improve personal safety in real-world situations. Safety during travel and emergencies is a serious concern, and I wanted to work on a solution that combines hardware and software to provide immediate assistance when needed.
Purpose
The purpose of this project was to design a portable safety system that can send real-time alerts and location information during emergencies. By using sensors, GPS, and wireless communication, the goal was to create a simple and reliable device that could be easily integrated into everyday use.
Challenges
One of the main challenges was integrating multiple hardware components while ensuring reliable communication and low power consumption. Handling real-time GPS data, maintaining stable Bluetooth connectivity, and designing the system to be user-friendly while remaining cost-effective required careful planning and testing.
Booklib Library Management Website
Web app using Django, JavaScript, HTML, CSS, and Python for book management, user authentication, and personalized recommendations.
GitHub Repository
Why I built this
I built this project as part of the HackFest hackathon organized by the CSI VIT-AP chapter. The event focused on learning how to build websites from scratch, and this project gave me the opportunity to apply those concepts in a practical, time-bound environment while collaborating and problem solving under hackathon conditions.
Purpose
The purpose of this project was to create an interactive web application to help manage books in a library. The goal was to simplify basic library operations such as book tracking and user interaction while improving the overall usability compared to manual or non-interactive systems.
Challenges
Building the website from scratch within a limited time frame was a major challenge. Designing an interactive interface, structuring data logically, and ensuring smooth navigation required quick decision-making and constant iteration. Balancing functionality and simplicity during a hackathon was another key challenge.
AI Story Generator | IBM Generative AI Certification Project
Built using GPT-2 via Hugging Face, Flask backend, and React + TypeScript frontend. Generates creative stories from user prompts in real time.
GitHub Repository
Why I built this
I built this project as part of my IBM Generative AI certification to practically apply what I was learning in the course. Instead of limiting myself to theoretical concepts, I wanted to prove my understanding of generative AI by building a complete, working application. Creating an AI that could generate new stories from user prompts felt like a fun and meaningful way to explore how language models behave in real scenarios.
Purpose
The main purpose of this project was to gain hands-on experience with generative AI and understand how text generation models work beyond examples and notebooks. I wanted to learn how to deploy a model, control its outputs, and integrate it into a real-time web application. This project helped me connect AI concepts such as prompt design, sampling strategies, and model inference with practical software development.
Challenges
One major challenge was dealing with hallucinations in the generated text, where the model would repeat the same word or phrase multiple times. This required experimenting with generation parameters such as temperature and sampling methods to reduce repetition while maintaining creativity. Another challenge was identifying a suitable model that could generate coherent stories while remaining lightweight enough to run efficiently. Solving these issues improved both the quality of the generated stories and my understanding of how generative models behave in practice.
Portfolio Website | Personal Project
Personal portfolio website to showcase projects, certifications, and skills.
View Portfolio
GitHub Repository
Why I built this
I built this portfolio website to present my projects, skills, and experiences in a structured and accessible way. Rather than relying on pre-made templates, I wanted full control over the design and functionality so that the portfolio itself could serve as a demonstration of my frontend development skills.
Purpose
The purpose of this portfolio is to clearly showcase my academic work, technical projects, and research interests to recruiters and collaborators. It also serves as a space to continuously improve the site by adding new features and refining the user experience, while keeping the website lightweight and fast using GitHub Pages.
Challenges
One challenge was organizing a large amount of information without overwhelming the user. I had to carefully structure sections, balance content density, and ensure responsiveness across devices. Implementing interactive features such as smooth navigation, dark and light mode, and expandable project details while maintaining performance was another key challenge.
AGSA – Automated Government Service Agent | Multi-Language
Developed multilingual, voice-enabled AI chatbot using Google Gemini API to simplify citizen access to government services. Supports 11+ Indian languages, integrates with DigiLocker, UMANG, CPGRAMS, Bhashini.
GitHub Repository
Why I built this
I built AGSA as part of the IndiaAI Impact Generative AI Hackathon organized by IISc CNI and IBM Research. The hackathon focused on using generative AI to solve real-world problems, and this project gave me the opportunity to work on a socially impactful use case involving access to government services.
Purpose
The purpose of AGSA was to design a multilingual, voice-enabled AI assistant that could help users navigate government services more easily. The goal was to reduce language barriers and complexity by providing conversational guidance while integrating with existing digital governance platforms.
Challenges
Handling multilingual input and maintaining accurate, context-aware responses across different user queries was a significant challenge. Designing effective prompts and managing integrations with external platforms required careful planning. Balancing social impact, system complexity, and usability within the hackathon timeline was another key challenge.
Automated Microplastic Detector | VIT-AP University (Jan'26)
Developed a Deep Learning-based image segmentation pipeline to automate the identification and quantification of microplastic contaminants in environmental samples.
GitHub Repository
Why I built this
As part of my Engineering Clinics project, I aimed to solve a critical bottleneck in aquatic research: the manual identification of microplastic pollutants. Traditional monitoring methods in environmental science are incredibly time-consuming and prone to human error, as they rely on researchers manually counting microscopic particles. I developed this automated detection pipeline to transform hours of manual lab work into sub-second digital analysis, providing an objective and scalable way to measure plastic concentrations in water bodies and protect aquatic health.
Purpose
Developed as part of an Engineering Clinics initiative, this project addresses the global challenge of micro-pollutant monitoring. Manual identification of synthetic particles in water samples is a slow, subjective process that hinders large-scale research. I built this automated detection pipeline to transform hours of laboratory microscopy into sub-second digital analysis. By creating an objective, machine-driven metric, the system provides a scalable tool for wastewater facilities and marine biologists to accurately track pollution levels and protect aquatic health.
Challenges
The primary challenge involved the high-fidelity detection of microscopic fibers, which are often so thin they can be "lost" or blurred during standard image processing. I solved this by implementing an architecture that preserves high-resolution spatial details, ensuring the geometric properties of fine synthetic threads are maintained. To handle "noise"—such as organic debris and air bubbles that mimic plastic—I engineered a preprocessing pipeline to normalize pixel intensities, allowing the model to distinguish synthetic polymers from natural matter. Finally, to overcome the extreme data imbalance where plastic occupies only a tiny fraction of the image, I utilized pixel-wise overlap metrics to ensure the system prioritized precise pollutant identification over the empty background.
Achievements & Hobbies
- Gold Medal & Diploma of Distinction - International Youth & Child Art Exhibition, 2012
- First Place - Indian Oil Drawing Competition, 2013
- State-Level Painting Competition - Energy Conservation Campaign, 2013
- Global Mathematical Talent Probe - 78.9 percentile, IFSE Institute, 2013–14
- Third Place (Early Bird) - Digital India Poster Contest, 2015
- Summer Theatre Workshop - National School of Drama, 2014
- Great India Hack-Fest 2k24 - Selected for Round 2 (CSI Chapter, VIT-AP)
Positions of Responsibility
- Marketing Lead - Bongojo (VIT-AP Bengali Association), Jan’25–Mar'26
- Research Team Member - SEDS VIT-AP (part of SEDS India), Oct’23–Present
- Documentation Member - Bongojo, Nov’23–Jan’25
- Marketing Team Member - Geeks For Geeks VIT-AP, Nov’23–Jan’24
- Social Media Co-Lead - North-East Association, VIT-AP, Sep’25–Mar'26
Contact
Email: saha92873@gmail.com
GitHub: TheSlayer7
LinkedIn: Ayush-Saha