Mithil Baria

AI/LLM Engineer and Full-Stack Software Developer with 5+ years of experience building secure, scalable applications and AI-driven systems across domains. Expert in designing backend architecture using Java, Spring Boot, and RESTful APIs, and deploying ML solutions leveraging PyTorch, Transformers, and RAG frameworks. Demonstrated success in driving 25–50% improvements in performance, security, and delivery speed. Published researcher and hackathon winner passionate about bridging intelligent systems with real-world enterprise use cases. Open to roles in software engineering, AI/ML, and tech consulting.

Code Reusability with RAG: Developed RAG system retrieving code snippets from structured corpus based on user input. Integrated with Streamlit UI; boosted software delivery and reduced redundant coding. 

https://github.com/Mithil21/MemoryMatrix

Fake Review Detection (Published): Built web app using RoBERTa to flag suspicious reviews on Amazon. Hosted full-stack system (Flask + Angular). IJMTST Paper

URL Shortener with Docker: Created a containerized microservice for URL shortening using Java and Spring Boot. Deployed using Docker for isolated and scalable environments. 

https://github.com/Mithil21/URLShortnerOnDocker

Positivity Generator (GPT-2 LoRA): Fine-tuned GPT-2 using LoRA/QLoRA for delivering personalized motivational messages.

Sarcasm & Fake News Detection: Designed NLP models (RNN/LSTM, BERT) achieving 80–85% accuracy for classification tasks.

Gujarati OCR Engine: Built ML model for handwritten Gujarati script recognition, achieving 82% accuracy for regional digitization.

• Winner, Code Games Hackathon, 2024
• Certified in Java, Python, JavaScript (Grasshopper)
• Paper Published: Fake Review Detection using RoBERTa, IJMTST
• Developed & Deployed AI tools using RAG, Transformers, and custom fine-tuning techniques

• National Level Cricket Player
• Trekking Leader
• Book Author
• Enthusiast in Swimming and Boxing