Adhithya.dev
I build intelligent web apps and data platforms that merge AI, analytics, and modern development. I’m skilled at creating machine learning prototypes, end-to-end software systems, and clear data visualizations that drive business decisions.
I'm Final-year B.Tech student in Computer Science and Business Systemsstudent with a strong passion for building intelligent solutions that combine technology, creativity, and real-world impact. My journey so far has been a blend of academics, hands-on development, and content creation. I specialize in areas like machine learning, full-stack development, and business-driven problem solving, with projects that include an AI Based Skin Diagnostic System using MobileNetV2 and metadata fusion on the HAM10000 dataset, a Hospital Appointment System developed during my internship at Retech Solutions using Java and MySQL, and an Event Management Website designed with PHP, JavaScript, and SQL.
Beyond development, I enjoy exploring data-driven insights through Power BI dashboards, working with APIs, and applying transfer learning strategies for healthcare and analytics applications. I actively participate in hackathons and innovation challenges such as the MSME Idea Hackathon and contribute to my Institution’s Innovation Council. Alongside coding, I create and share knowledge through blogs, workshops, talks, and podcasts, helping peers learn technical and career-related skills.
I see myself not just as a developer, but as a problem solver and innovator—someone who thrives at the intersection of technology, research, and business systems. My goal is to keep learning, contribute meaningfully through impactful projects, and grow into a professional who builds solutions that are both technically sound and valuable to society.
Built a transfer-learning pipeline using MobileNetV2 and combined image probabilities with metadata models (Logistic Regression / Random Forest / XGBoost) to improve diagnostic accuracy. Deployed a demo using Streamlit.
An event planning platform with event creation, vendor selection, and comparison. Includes a roadmap to add Dialogflow chatbot for assisted planning.
Developed a sentiment analysis project using NLP and executing it on Google Colab to classify text as positive, negative, or neutral.
A stegnography apprach that enables to hide data either text or pdf that could be hide into a image.
Interactive Power BI dashboard to track consumption trends, detect anomalies, and provide insights for cost optimization.
Full-stack Java application for appointment management with role-based access. Implemented prepared statements, session management, and Postman-tested endpoints.
• AI Skin Diagnostic System using MobileNetV2 (ongoing)
Research focused on building a deep learning model for early skin disease detection, preparing for publication as a research paper.
• Women Safety Analytics Platform (2024) - Smart india hackathon
Developed a data-driven solution for analyzing safety patterns and enhancing women’s safety through AI-based monitoring and predictive analytics.
• AI-based Early Disease Detection Platform (2024) -MSME Idea Hackathon 4.0
Proposed an intelligent healthcare solution leveraging machine learning to detect diseases at an early stage, improving preventive care and patient outcomes.
provide valuable materials and resources for peers and provided Git/GitHub guidance.
AI models can accelerate early screening for diseases when deployed responsibly. In this post I explain transfer learning with MobileNetV2, dataset nuances (HAM10000), class imbalance mitigation, and the importance of interpretability and clinical validation.
Introduction — Early disease detection is a field with high potential; ML can assist dermatologists by prioritizing suspicious lesions for review. I used MobileNetV2 pre-trained on ImageNet and fine-tuned with augmentations to tackle the HAM10000 dataset's skewed classes.
Approach — Data cleaning, stratified k-fold, calibration, metadata fusion (age, sex, lesion location) via XGBoost stacking. While accuracy improved, careful error analysis revealed biases — a reminder to treat ML as an assistant, not a diagnosis tool.
Conclusion — ML systems must be validated with real clinical workflows, and ethics & privacy must be central.
Studying CSBS taught me to connect technical solutions to business outcomes. I outline frameworks for thinking in product terms, simple metrics to measure value, and how to present technical work to non-technical stakeholders.
When building an ML model, ask: what business question does it answer? Be explicit about KPIs (time saved, accuracy improvements) and show prototypes early. Cross-functional communication and concise one-page documentation help projects win buy-in.
This post shares a daily practice routine, how to track progress, and tips for staying motivated. I include a 12-week schedule for problem-solving and resources to follow.
Key habits: short daily goals, review & annotate solutions, teach peers, and use spaced repetition. I also talk about using GitHub to store your problem solutions and building mini-projects to apply new concepts.
Presented methodology, model performance, limitations, and ethical considerations. Slide deck includes architecture, dataset exploration, and demo screenshots.
Hands-on session covering HTML/CSS/JS, GitHub workflows, and a small MERN overview. Followed by Q&A and 1:1 mentoring sessions.
Pitched solutions for AI early disease prediction and early assist support using AI and digital marketplaces. Focused on wellbeing and early action.
Show notes: Discussion about model choice (MobileNetV2), data challenges, deployment on Streamlit, and future roadmap.
Show notes: Product decisions, vendor comparisons, and adding conversational UX with Dialogflow.
Show notes: Practical tips to stay consistent, scheduling practice time, and using projects to demonstrate skills.
I’d love to collaborate, discuss opportunities, or share ideas. Feel free to reach out through any platform below.