Pras Bairwa | CS @ UIUC | James Scholar
ML Systems · Research · Engineering

Pras Bairwa

AI Systems & Machine Learning Engineer

Computer Science @ UIUC (4.0 GPA)

Building high-performance retrieval systems, multimodal AI pipelines, and computer vision models.

View Work CV GitHub LinkedIn

Core technical work

Industrial RAG system for digital twin data

Built a retrieval-augmented generation system from scratch using LangChain to query industrial digital twin data at Kongsberg Digital. Implemented parallel retrievers and a multimodal retrieval pipeline for engineering documentation and operational data, and integrated Corrective RAG (CRAG) and Reflexion-style feedback loops to improve response accuracy. Achieved a 70% reduction in query latency through concurrent retrieval and optimized orchestration.

Computer vision for biomedical imaging

Developed deep learning models for label-free SLIM microscopy segmentation using CNNs, YOLO, and Cellpose. The approach enables automated analysis of cellular structures without manual labeling, reducing annotation cost and enabling scalable analysis of microscopy data for traumatic brain injury research.

Computer architecture systems

Contributed to the maintenance of SpimBot, a MIPS-based autonomous agent simulator used in UIUC’s computer architecture course (CS233). Served as a course assistant, helping students understand instruction execution, pipelining, and memory hierarchy concepts during office hours.

Designing retrieval workflows for real-world data.

Industrial AI systems

At Kongsberg Digital, I worked on retrieval based systems for industrial digital twin data, focusing on orchestration, multimodal retrieval, response correction, and latency-sensitive system design. The work emphasized practical AI engineering: building systems that can handle noisy documentation, operational context, and multiple model backends.

LangChain · RAG · Multimodal Retrieval · CRAG · Reflexion · LLM Orchestration

Engineering and teaching

Beyond AI systems work, I’ve also worked in teaching-oriented engineering roles at UIUC. In CS233, I supported a simulator-based computer architecture course through SpimBot maintenance and office hours. In CS124 Honors, I mentored students on convolutional neural network fundamentals and their project analyzing galaxy morphology through CNNs. Their project ended up in the Hall of Fame. Additionally, I authored course material to facilitate introduction to computer vision.

C++ · MIPS · Computer Architecture · Debugging · Mentoring

See here for Hall of Fame →

See here for Keras Course Material →

Machine Learning Research

Label-free tissue analysis

My research explores computer vision methods for biomedical imaging, particularly in settings where labeled data is limited or expensive to obtain. In the Cellular Neuroscience Imaging Lab, I developed models for label-free segmentation and automated analysis of microscopy data, with an emphasis on scalable, research useful pipelines. My research is under the supervision of Dr. Catherine Best-Popescu.

Computer Vision · Biomedical Imaging · Deep Learning

See here for Dr. Catherine Best-Popescu →

Senior thesis

I am completing my senior thesis under the supervision of Dr. Pablo Robles-Granda, studying retrieval-augmented generation systems and methods for improving reliability and reasoning in LLM-based pipelines.

The work investigates techniques for retrieval orchestration, evaluation, and feedback mechanisms in retrieval-augmented models, building on my broader interest in scalable AI systems and information retrieval.

RAG Systems · Information Retrieval · LLM Systems · Research Engineering

See here for Dr. Pablo Robles Granda →

Selected technical work

Transformer from scratch

Implemented a full encoder–decoder Transformer architecture in PyTorch, including custom multi-head attention, positional encoding, and attention masking. The project reproduces the core architecture of modern sequence models while exposing the mechanics of attention, token embeddings, and gradient flow in deep learning systems.

View code →

Local retrieval-augmented generation system

Built a local RAG pipeline using LangChain, ChromaDB, and Ollama embeddings to enable semantic document retrieval and LLM-based question answering. Implemented document chunking, embedding indexing, and retrieval pipelines for queryable knowledge bases.

View code →

Federated facial detection

Designed a privacy-preserving federated learning system in Rust that trains computer vision models across distributed clients using federated averaging. The system simulates decentralized training while keeping raw data local, demonstrating approaches for privacy-aware machine learning.

View code →

Technical toolkit.

Languages & Systems

Python SQL Java C++ C Rust Verilog MIPS Assembly

Machine Learning

Deep Learning Computer Vision Natural Language Processing Retrieval Augmented Generation Multimodal AI Federated Learning

Frameworks & Libraries

PyTorch TensorFlow Scikit-learn LangChain OpenCV

Tools & Infrastructure

Docker Git Linux Azure DevOps

Awards

University of Illinois Urbana-Champaign

B.S. in Computer Science at the Grainger College of Engineering.

Maintaining a 4.0 GPA while pursuing research in machine learning and AI systems.

Relevant Coursework

Data Structures · Computer Architecture · Probability & Statistics for Computer Science · Linear Algebra · Numerical Methods · Discrete Structures · Algorithms · Text Information Systems · Database Systems · Differential Equations

Honors & Awards

URSA Best Overall Presentation Award

James Scholar and Dean’s List (all semesters).

Scholarships

Illinois Engineering Achievement Scholarship

Grainger College of Engineering Outstanding Scholarship