+ {statusMessage} +
+ )}Expertise
-Full Stack Web Development
-I have built a diverse array of web applications from scratch using modern technologies such as React and Flask. I have a strong proficiency in the SDLC process and frontend + backend development.
-Expertise
+Data Science & Analytics
++ I transform raw data into actionable insights through data pipelines, advanced analytics dashboards, + and ETL workflows, enabling organizations to uncover patterns, automate processes, and make confident + data-driven decisions. From building executive dashboards to extracting insights from complex datasets, + I excel at turning data into strategic value. +
+{/* This adds a line break */} + {dataScienceSkills.map((skill, index) => ( +
DevOps & Automation
-Once the application is built, I help clients set up DevOps testing, CI/CD pipelines, and deployment automation to support the successful Go-Live.
-Machine Learning & Computer Vision
++ I specialize in developing and deploying machine learning models to solve real-world problems, + including clustering, sentiment analysis, and object detection. In computer vision, + I’ve delivered solutions like advanced image segmentation and CycleGAN-based domain adaptation. + By optimizing models and tuning hyperparameters, I ensure accuracy and scalability in every implementation. +
+{/* This adds a line break */} + {machineLearningSkills.map((skill, index) => ( +
GenAI & LLM
-Stay relevant in the market by leveraging the latest AI models in your projects. I have professional experience building enterprise grade GenAI-enabled solutions to empower intelligent decision making.
-Generative AI & NLP Research
++ I leverage state-of-the-art AI technologies to fine-tune LLMs for enterprise challenges, design + Retrieval-Augmented Generation (RAG) systems for intelligent data retrieval, and implement NLP solutions + for extracting insights from unstructured text. Expertise with tools like LangChain and Hugging + Face enables me to build scalable, production-ready systems that drive innovation and efficiency. +
+{/* This adds a line break */} + {generativeAISkills.map((skill, index) => ( +
+ Yuji Sato
-Full Stack Engineer
+Aradhya Dhruv
+Research Associate at ATMRI (NTU) | MSc Artificial Intelligence | AI Researcher in Aviation Systems, Speech AI & LLM-driven Decision Support | Former Trainee Decision Scientist at Mu Sigma | Enabled Data-driven strategy for a leading Fortune 500 company
Personal Projects
-Filmate AI
-Developed movie finder app with semantic search and sentiment analysis using OpenAI GPT-3.5 Turbo, Qdrant, React, and Flask.
-High Speed Chase
-Designed, developed, and launched a 3D multiplayer racing game with C# and Unity. This is available on Itch.io for gamers worldwide to enjoy.
-Astro Raiders
-Developed and released a 2D shooting game with C# and Unity. This project is hosted on the Itch.io public marketplace.
-Datum: Integrated Learning Platform
-This is an online educational platform that provides high-quality, data science-focused learning resources in the Japanese language. I created the entire platform from scratch using Ruby on Rails.
-WeManage: Real Estate Asset Management
-This mobile application allows realtors in Japan to securely manage their property information and view future income predictions. This app is built with Ruby on Rails and JavaScript.
-COVID-19 Case Management
-Built official charts for COVID/vaccination tracking for an educational institution using JavaScript and the Google Sheets API v4. The dashboard served the university's leadership in their decision-making processes.
-Multiple Regression Property Analysis
-Analyzed the real estate market in Japan and predicted property prices by implementing statistical methods such as OLS and multi-regression analysis. This project leveraged Python and various libraries such as Pandas, NumPy, Matplotlib, and Scikit-Learn.
-Programs of Study
-Designed and developed a custom component for a CMS-based platform (e.g., 'Brightspot') using Java, Handlebars, and LESS. University students can find their majors of interest through this module.
-Transfer Evaluation Matrix
-Created an interactive CSV table generator with Java, Handlebars, and LESS. This project helps transfer students to quickly identify eligible credits.
-Submeowrine
-Developed and released an Android mobile application using Java and Android Studio that runs a 2D shooting game.
+ return ( +Personal Projects
+LLM-powered Air Traffic Scenario Generation
++ Built a chat-based framework that uses LLMs plus Graph RAG on Neo4j to ingest ATC scenario PDFs and generate new, + diverse scenarios. The system injects controlled conflict cases, stores them back in the graph for validation and reuse, + and supports structured JSON and XML exports for simulation workflows. +
+LLM-augmented RL for Aviation Decision Support
++ Developed an LLM-guided PPO framework for aircraft conflict resolution and aviation decision support. Studied how LLM + guidance affects policy stability and reward convergence under complex operational dynamics. +
+Semantic Guided Image Augmentation (SGID)
++ Implemented a novel image augmentation method for diffusion models leveraging Stable Diffusion, BLIP, and CLIP. + Developed an algorithm to maintain semantic consistency by generating captions and prompts, ensuring diversity + and improving image classification performance. Applied Gaussian noise for controlled augmentation. +
+Image Classification on CIFAR-100
++ Enhanced prediction accuracy on the CIFAR-100 dataset by applying advanced regularization techniques, including + cosine annealing and weight decay. Employed data augmentation strategies to prevent overfitting and improve + generalization in deep learning models. +
+Sentiment Analysis with Sequence Models
++ Conducted an in-depth study of neural network architectures for sentiment analysis, including feedforward + networks, CNNs, LSTMs, and BiLSTMs. Evaluated performance using F1-score, precision, and recall, and analyzed + the impact of pre-trained embeddings and various optimization strategies. +
+Seq-to-Seq Models for Machine Translation
++ Implemented and compared Seq2Seq architectures, including LSTM, BiLSTM, and Transformer, for machine + translation tasks. Fine-tuned hyperparameters and evaluated models using BLEU scores to optimize + translation quality across languages. +
+Newsletter Generator with CrewAI
++ Developed an automated newsletter generation system using CrewAI. Designed an intuitive GUI for users to + input topics and personal messages. Implemented autonomous AI agents to summarize and format content, delivering + newsletters as downloadable HTML files. +
+Image-to-Image Translation with CycleGANs
++ Implemented CycleGANs for unpaired image-to-image translation, ensuring realistic domain adaptation + using adversarial training and cycle consistency loss. Demonstrated successful translation between + distinct image domains without paired data. +
+Object Detection with YOLO-NAS
++ Built a real-time object detection interface using YOLO-NAS. Developed a Streamlit-based application + allowing users to upload images and videos and view detection results with high accuracy and efficiency. +
+Chat with PDFs Using LangChain
++ Developed an application enabling users to query and retrieve content from PDFs using LangChain and RAG. + Integrated LLMs for accurate and contextualized responses, enhancing document comprehension. +
+Stock Market Analysis with PEFT
++ Automated financial analysis by integrating Parameter-Efficient Fine-Tuning (PEFT) and QLoRA for LLM optimization. + Incorporated sentiment analysis and summarization to provide actionable insights for stock market trends. +
+Q-Learning for Cliff Box Pushing Task
++ Implemented Q-learning to solve a customized grid-world problem where an agent navigates, pushes a box, + and avoids cliffs to reach a goal. Designed and compared exploration strategies including Epsilon Decay, UCB, + and Standard Q-learning, achieving a 100% success rate. +
+Career History
Technology Consultant
-Dallas, TX
+Research Associate
+Air Traffic Management Research Institute (ATMRI), Singapore
- Full-stack Web Development, GenAI/LLM, Project Management, Business Development + I develop and experiment with AI systems that improve multi-accent air traffic control speech recognition and information extraction, using deep learning, large language models, and knowledge graphs to support safer and more reliable aviation operations.
+ + {/* Stealth Startup */} +AI Researcher
+Pursoo (Startup), Singapore
++ Developed an AI-powered candidate recommendation platform with graph database similarity search, PII handling via Presidio, and advanced resume formatting using LLMs. +
+ + + {/* NTU Student Assistant */} +Student Assistant
+Nanyang Technological University, Singapore
++ Developed a multi-agent RAG chatbot for course recommendations, integrating FAISS, Sentence Transformers, and LLaMA 3.1 models. +
+ + + {/* Nanoveu */} +AI Engineer
+Nanoveu, Singapore
++ Reduced latency for 2D-to-3D conversion with Mask R-CNN and hyperparameter tuning on Deeplab V3 for segmentation quality. +
+ + + {/* Mu Sigma */}Full Stack Engineer
-Laie, HI
+Decision Scientist
+Mu Sigma, India
- Frontend Development, Backend Development, User Experience, Team Leading + Designed executive dashboards, built bankruptcy prediction models, and developed sentiment analysis tools, enabling strategic insights for a U.S. energy company.
+ + {/* Malviya National Institute of Technology */}Staff Engineer Intern
-Laie, HI
+Machine Learning Intern
+MNIT, India
- Full-stack Development, API Development, User Experience + Developed classification models using TF-IDF and BoW for retail analytics, deriving insights from customer reviews and demographics.
+ + {/* NUS */}Data Analyst Intern
-Tokyo, Japan
+Machine Learning Intern
+National University of Singapore
- Automation, Data Governance, Statistical Analysis + Built IoT systems for real-time analysis of RFID data using Azure IoT Hub and Power BI to optimize bus passenger congestion.