About
I am pursuing a Double Major in Computer Science & Mathematics at WashU with a GPA of 3.95. My engineering philosophy sits at the intersection of rigorous distributed systems and pragmatic AI application.
I don't just tune models; I build the infrastructure that makes them viable in production. From scaling Kafka-backed async processing to support 5k+ concurrent sessions, to optimizing RAG pipelines that cut operational costs by 27%, I focus on impact.
When I'm not deploying Kubernetes clusters or creating Llama-Vision scanners, I'm competing in math benchmarks (AIME Qualifier), singing, or exploring Jazz Studies.
Experience
-
May — Aug 2025 Engineered full-stack LLM solutions and scalable backend infrastructure. Built a hybrid ATS scoring engine combining TF-IDF with LLM rewrite guidance. Scaled Node/Express systems using Kafka-backed async processing to handle 5.1k concurrent sessions. Implemented strict circuit-style fallbacks and timeouts to harden LLM APIs.
- Postgres
- Kafka
- Redis
- Elasticsearch
- AWS
-
May — Aug 2024 Owned the end-to-end RAG platform for offline news analysis. Integrated RouteLLM to route queries to local models, reducing cloud inference costs by ~38%. Built a self-serve admin dashboard (MongoDB/SSO) for retrieval tuning.
- LangChain
- MongoDB
- RAG
- Azure
Selected Projects
-
NL Video Scanner (Llama Stack Finalist)
Privacy-preserving scanner running at 1 FPS using Llama-Vision + OpenCV. Achieved 85%+ Hit@5 with zero data egress. Built local retrieval with LlamaIndex and optimized OpenCV preprocessing for low-light scenes.
- Llama-Vision
- LlamaIndex
- NoSQL
- OpenCV
- FastAPI
- Docker
-
E-Commerce Recommendation Engine
Two-stage recommender (recall → ranking) processing 10M+ interactions. Served a Two-Tower DNN via TorchServe on Kubernetes (p99 9ms @ 1.2k RPS). Implemented Spark Structured Streaming for real-time feature caching.
- PyTorch
- Kubernetes
- Spark
- Redis
-
Distributed RPC Framework
Designed a low-latency RPC core in Spring Boot + Vert.x. Reached <5ms median latency at 5k+ RPS. Implemented Etcd-based service discovery and fault-tolerant failover, validated on a multi-node Linux cluster.
- Distributed Systems
- HPC
- Spring Boot
- Etcd
- Vert.x
-
ML Comparative Analysis
Conducted a comparative study of supervised learning models (SVM, Logistic Regression, Random Forest) on physicochemical datasets. Performed rigorous EDA to identify key correlations (e.g., alcohol vs. quality). Achieved 87.2% accuracy with Random Forest while optimizing SVM for Macro F1-score balance across minority classes.
- Scikit-Learn
- Pandas
- Matplotlib
-
SoloSync Travel App (iOS)
Built a social travel platform bridging isolation for solo travelers. Integrated MapKit and Google Places for interactive route planning and real-time annotation sharing. Engineered a scalable Node.js/Express backend on AWS EC2 with MariaDB for reliable data persistence.
- SwiftUI
- Node.js
- AWS
- MapKit
- Google Places
-
Hollow Archer (2D Action Platformer)
Engineered a Metroidvania-style platformer in Unity. Architected a hierarchical Enemy AI system using C# inheritance (Patrol/Chase logic). Implemented kinetic combat physics, custom health data structures, and optimized WebGL memory management for browser play.
- Unity
- C#
- WebGL
- Game Design
