I'm Wesley Deklich, a sophomore at the University of Illinois Urbana-Champaign, pursuing a major in Computer Science + Economics with a statistics minor. My academic journey is driven by a passion for leveraging artificial intelligence and machine learning to solve complex real-world challenges.
With a strong foundation in AI/ML research and engineering, I specialize in natural language processing, neural search systems, and large-scale data analysis. My experience spans from implementing cutting-edge RAG pipelines and ontology-driven search to developing production-ready machine learning systems that serve millions of users.
I am deeply fascinated by the intersection of AI research and practical engineering, focusing on optimizing model performance, building scalable ML infrastructure, and advancing the state-of-the-art in natural language understanding and information retrieval systems.
→ Implemented ontology-driven neural search for RAG pipelines by integrating structured entity hierarchies with dense embeddings from sentence-transformers, increasing query matching accuracy by 18% in internal benchmarks
→ Developed unsupervised clustering workflows using HAC, UMAP, and HDBSCAN to organize large-scale enterprise knowledge base entries, improving retrieval precision by 22% and reducing noise in results
→ Optimized FAISS ANN indices (IVF-PQ, HNSW) for low-latency, high-recall semantic search on multi-million vector datasets, cutting average retrieval time from 120ms to 45ms without loss in recall
→ Identified critical flaws in Bayesian Classification algorithms and meteorite pairing guidelines by analyzing 10,000+ historical data points using Python, TensorFlow, and NumPy, improving model precision by 15%
→ Developed a new meteorite pairing framework based on physical attributes, geographic proximity, and orbital history, increasing classification accuracy by 20% across 5,000+ samples
→ Presented findings at AGU Fall Meeting, engaging over 25,000 attendees globally via a live presentation and Q&A
→ Worked closely with a team to design and implement cloud-based communication solutions by optimizing server architecture and enhancing data compression, significantly improving video services for over 1M global users
→ Assisted in analyzing real-time video usage data using Python and Tableau, successfully identifying key areas of improvement to reduce video latency, and streamlined the data processing workflow
→ Leveraged Tableau and D3.js to design dynamic, scalable, and interactive visualizations, incorporating real-time data updates and intuitive UI/UX principles, which boosted total user engagement by 20%
GPA: 4.0
→ Developed and integrated a Natural Language Processing (NLP) pipeline using Python libraries such as NLTK and SpaCy to analyze investor sentiment from unstructured text data, enabling real-time sentiment visualization
→ Worked closely with the database team to design a MongoDB schema for efficient data storage and retrieval, optimizing overall performance for handling over 500,000 data points effectively
→ Integrated Axios for seamless API communication, ensuring a 99.9% uptime and reducing data retrieval time by 30%, while handling over 10,000 API requests daily with optimized response times
→ Developed a cross-platform note-taking app using Electron and TypeScript, implementing state management with Jotai and optimizing build processes for seamless deployment across Windows, macOS, and Linux
→ Engineered core functionalities including a Markdown editor with real-time preview and data persistence through Electron's IPC, enabling reliable and user-friendly note-taking capabilities
→ Integrated TensorFlow.js to develop a smart recommendation system that suggests relevant tags, categories, and even related notes, improving user productivity, organization, and content discovery
→ Developed an AI-based classification model to distinguish between meteorites, craters, and ordinary landscapes using computer vision techniques and convolutional neural networks
→ Leveraged TensorRT & PyTorch to optimize real-time inference for deployment on NVIDIA Jetson Nano and other edge computing hardware
→ Implemented a convolutional neural network (CNN) trained on high-resolution geospatial imagery to identify key morphological differences between meteorites and terrain
→ Designed an automated drone-based scanning system, enabling researchers to survey large areas efficiently and detect potential meteorite sites
→ Immersed in hands-on experiences, lectures, and company visits across Silicon Valley
→ Generated a new business idea and assessed its profit potential, built innovative business models, and iterated toward product-market fit
→ Presented "Tessero" in front of Wharton faculty, alumni, and successful entrepreneurs
→ Won first place overall, nominated "Most likely to be the next Steve Jobs"
→ Founded a nonprofit organization by filing LLC status forms and bylaws, aimed at monitoring sea turtle birth nests, tracking migration patterns, and organizing large-scale beach cleanup initiatives
→ Successfully raised $13,000+ in donations, allocating funds to large conservation groups and aquariums to support preservation efforts
→ Developed the foundation's website using HTML5 and SCSS, optimizing it for mobile responsiveness and accessibility
→ Established a Kindle reselling business leveraging Amazon FBA (Fulfillment by Amazon), handling logistics, inventory management, and automated order fulfillment
→ Developed a full-stack web application for Kindle sales using HTML, CSS, JavaScript, and Swift, integrating real-time price tracking APIs to optimize sales strategies
→ Developed an enterprise-grade fintech platform to facilitate secure decentralized transactions
→ Designed and deployed a Polygon (ERC20) blockchain-based token, ensuring compliance with smart contract best practices using Solidity
→ Integrated with UniSwap (decentralized exchange) to enable token liquidity and peer-to-peer transactions
→ Conducted data-driven research on educational inequalities in low-income communities by analyzing student performance metrics and resource allocation disparities
→ Published findings on ERIC.GOV (2021), contributing to ongoing educational policy discussions