A picture of Nikola

Computer Science @ Texas A&M University

|

About

Background

I'm a fourth-year Computer Science student at Texas A&M University, where I also work as an undergraduate researcher. My experience includes fine-tuning open-source LLMs with LoRA and integrating retrieval-augmented generation into a Grid Resilience Assistant. I co-authored an IEEE paper on visualizing historical weather and renewable energy trends. I've built distributed systems using gRPC and RabbitMQ and implemented core operating system components in C, including virtual memory, threading, and a basic file system. I’ve also developed backend services with Flask, FastAPI, Express.js, and PostgreSQL, and built web interfaces with React and Next.js. I'm always exploring new technologies and focused on optimizing performance across the systems I design.

Work Experience

Undergraduate Researcher

Texas A&M University, Department of Electrical and Computer Engineering

May 2024 - Present

  • Developed a Grid Resilience Assistant using a LoRA-fine-tuned LLM combined with Retrieval-Augmented Generation (RAG).
  • Reduced RAG context by 90% and cut latency via a FAISS vector DB, with minimal accuracy loss.
  • Co-authored an IEEE paper on visualizing weather & renewable-energy trends.
  • Cut an 80-year weather-analysis script’s runtime by 77% via parallel k-NN.
  • Implemented MILP-based power-scheduling models in Python (Pyomo & GLPK).

Software Engineer Intern

Teamup

May 2024 - August 2024

  • Developed a Next.js + Spring Boot full-stack app in an Agile team.
  • Built a React feedback UI with 5-star ratings & dynamic surveys.
  • Contributed to a mission-driven tutoring app advancing UN Sustainable Development Goals for education access.

Teaching Assistant

Texas A&M University, Department of Computer Science and Engineering

August 2023 - December 2023

  • Guided 50+ students through C++ labs and assignments.
  • Improved student outcomes through targeted support during office hours.
  • Led C++ workshops on dynamic memory, OOP, linked lists & recursion.

Skills

Hover over a skill for current proficiency

100%

90%

100%

80%

90%

75%

90%

100%

70%

90%

70%

80%

90%

70%

70%

80%

90%

80%

100%

100%

70%

60%

100%

90%

100%

Projects

Case Study 1 of 7: All-Hazards-AI RAG Platform

As an undergraduate researcher on the All-Hazards-AI RAG Platform at Texas A&M (Apr–Aug 2025), I built a system enabling any open-source LLM to chat, run user-uploaded Python scripts, and query local CSVs and PDFs. I architected a hot-reload FastAPI/WebSocket + gRPC stack to keep the GPU-resident model alive, reducing redeploy cycles by 95% and streaming first tokens in under 3 seconds. I built dual FAISS vector stores for CSV rows and PDF chunks, using normalized SentenceTransformer embeddings and IndexFlatIP to enable sub-100 ms k-NN retrieval. I also automated deployment with Bash scripts to install dependencies, compile protobufs, seed data, and launch the full stack on two NVIDIA H100 GPUs in under 30 seconds.
Visit at https://github.com/NikolaSlv/All-Hazards-AI.

Case Study 2 of 7: Panda Express Point of Sale (POS) System

As part of a team of 6, I contributed to the development of a scalable POS system for Panda Express, leveraging Python, React, and PostgreSQL with a cloud-hosted AWS database. The system featured real-time updates across Manager, Cashier, and Kitchen views, enabling seamless order processing and inventory management. I implemented WebSocket-based real-time updates, which reduced network usage by 60% and improved responsiveness by 80%, replacing inefficient polling methods. Throughout the project, we followed Agile practices, conducting weekly sprints and daily standups to ensure continuous improvement and delivery. The project repository is currently private; however, the final PowerPoint presentation is available via the following OneDrive link:
Visit at https://1drv.ms/p/c/6149f540eda847c0/ESlCYIxVTdpEkhPRsLwzKWABYQGgASH-PdwbN4rGTCrAaw?e=pKaF7L.

Case Study 3 of 7: SoundBytes, TAMUhack 2023 Winner

SoundBytes is an AI-driven platform designed to quickly and accurately summarize news articles and convert them into easy-to-understand audio clips. The project utilizes ChatGPT to extract relevant information and generate summaries, which are then transformed into audio through Microsoft Azure's Cognitive Services. Built using HTML, CSS, JavaScript, and Python with Flask, our team successfully navigated challenges related to API limitations and article formatting to create a dynamic, functional website.
Visit at https://devpost.com/software/soundbyte-q860bk.

Case Study 4 of 7: PicPredict, Google Quick, Draw! Image Classification

Led a team at a Texas A&M Datathon to develop a Convolutional Neural Network for classifying hand-drawn images into 15 major categories in real-time, achieving a top 5 placement and a 55% validation accuracy. Post-competition, independently redesigned the model architecture and trained it on a selected subset of the Google Quick, Draw! dataset, boosting validation accuracy to 86%.
Visit at https://github.com/NikolaSlv/PicPredict.

Case Study 5 of 7: Similar Day Analytics

This project is developed as part of a paper accepted for EnergyVis at VIS 2024, a leading workshop on energy data visualization. Similar Day Analytics utilizes the Frobenius Norm to identify days with comparable weather patterns. The project processes spatiotemporal data, such as temperature and wind speed, creating matrices to represent daily weather characteristics. The Frobenius Norm is then employed to calculate distances between these matrices, enabling the identification of similar days and enhancing meteorological analysis. The project is implemented in Python and incorporates matrix normalization along with k-nearest neighbors for accurate pattern matching.
Visit at https://github.com/NikolaSlv/SimDay-Analytics.

Case Study 6 of 7: E-Commerce Platform

Web-Store is an e-commerce website designed to manage orders for a wide range of products, including packaged foods and other consumer goods. It leverages Express.js, Node.js, and MongoDB, with a user interface built using CSS, Bootstrap, and JavaScript. The site follows the MVC architecture, enabling efficient data handling via REST API, and includes a secure user management system with encrypted passwords. This project emphasizes scalability, security, and a user-friendly design.
Visit at https://github.com/NikolaSlv/web-store.

Case Study 7 of 7: Store Management System for Desktop

The Java-SQL Management System was created to assist businesses in managing their operations efficiently by providing a robust system to handle their data needs. The idea behind the project is to offer a practical, easy-to-use application for small to medium-sized enterprises that require reliable data management but may not have the resources for expensive, complex systems. By integrating Java with MySQL, the system provides a cost-effective solution that can be customized and expanded as the business grows.
Visit at https://github.com/NikolaSlv/Java-SQL-Mgmt-System.

Contact

+1 (737) 895-2450

[email protected]