I'm an AI/ML Engineer specializing in building production-ready intelligent systems — RAG pipelines, NLP applications, and computer vision solutions. Currently pursuing B.Tech in Artificial Intelligence & Machine Learning and actively building toward a career in the Japanese tech industry.

My work sits at the intersection of language, intelligence, and real-world utility — whether it's bridging communication for the deaf community, detecting AI-generated content at scale, or building sentiment analysis tools for real business use.

• 🔭 Currently building: LLM-powered pipelines, MLOps workflows & AWS-ready deployments
• 🇯🇵 Japan-focused: JLPT N5 → targeting N4 by December 2026
• 🌐 Portfolio targets: Mercari, Abeja, PKSHA Technology, SmartNews
• 📚 Learning: LangChain advanced patterns · RAG optimization · AWS CLF-C02
• ⚡ Fun fact: I build AI tools that speak Japanese before I do

TensorFlow/Keras · MediaPipe · OpenCV · Streamlit

A real-time ASL recognition system trained on 87,000+ images across 29 classes. Uses MediaPipe for hand landmark detection feeding into a CNN classifier — achieving low-latency, camera-feed-ready translation output.

• Custom CNN architecture with fine-tuned MediaPipe hand keypoint input
• 29-class ASL alphabet coverage
• Deployable as a local Streamlit web app

🔵 LexBridge — RAG-Based Legal Document Simplifier

LangChain · ChromaDB · Groq · Streamlit

A RAG-powered application that lets users upload legal documents and ask questions in plain English, simplifying dense legal jargon into clear, grounded answers with cited source excerpts.

• Full RAG pipeline: chunking, embedding, vector retrieval, and LLM generation
• ChromaDB vector store with sentence-transformer embeddings
• Source-cited answers via Groq's LLaMA 3.1 for low-latency inference

HuggingFace RoBERTa · Flask · Python

A web application that classifies whether text is human-written or AI-generated, leveraging a fine-tuned RoBERTa transformer model. Built to address a real and growing problem in content authenticity.

• RoBERTa-based transformer inference pipeline
• Flask REST API backend with clean frontend
• Batch text analysis support

Python · Gradio · Excel

A local-first, privacy-focused sentiment analysis tool with batch text processing and Excel-optimized report generation. Built for businesses that need on-premise NLP without sending data to third-party APIs.

• Supports CSV/Excel batch input
• Clean Gradio dashboard UI
• Exportable analytics reports

N2 is the professional working proficiency target for Japanese tech companies.

• 🤝 Collaboration on RAG, NLP, or Computer Vision projects
• 💼 AI/ML internship or junior engineer roles — Japan-based or remote
• 📝 Technical writing on AI topics (publishing on Zenn.dev)
• 🔬 Research partnerships — cross-lingual NLP, sign language recognition