I am a Computer Science student at UC Irvine specializing in distributed data systems, infrastructure engineering, and applied AI/ML. I focus on taking complex machine learning prototypes and hardening them into scalable, high-throughput production architectures.
- 🎓 CS @ UC Irvine (Regents' Scholar, Dean's List), Graduating December 2026
- 🧱 Data & AI Systems: Designing enterprise-grade Medallion Lakehouses using Databricks, PySpark, Unity Catalog, and Delta Lake
- 🤖 Agentic Frameworks: Engineering stateful, multi-agent workflows using LangGraph and Model Context Protocol (MCP) servers
- ⚡ Core Systems: Building high-performance LLM routing, micro-batching, and semantic caching infrastructure in Go
- 💼 Industry Experience: Previously Software Engineering Intern @ Wells Fargo, shipping secure, high-coverage consumer APIs
- 🔬 Research & Leadership: Computer Vision Engineer @ UAV Forge, Proposal Fellow for NASA L'SPACE, and HCI researcher for accessibility development tooling @ UCI
- 👨🏫 Mentorship: Undergraduate Teaching Assistant / Learning Assistant for computer science courses @ UCI
- 🌐 Portfolio: mukundummadisetti.vercel.app
- 📫 Reach Me: Mukund.setti@gmail.com
|
LangGraph · Model Context Protocol (MCP) · RAG · YOLO · Vector Search · pgvector
|
Databricks (Serverless, Unity Catalog, Jobs API) · PySpark · Delta Lake · MLflow · Prometheus · Grafana · OAuth 2.0
|
Go · Python · TypeScript · Java · C/C++ · SQL · Next.js · Tailwind CSS
|
| Project | Description | Stack |
|---|---|---|
| 🧱 Astroturf: Regulatory Intelligence Platform | Databricks-native multi-agent Lakehouse platform engineered to track coordinated campaigns across federal dockets. Processes 211K+ public comments through a 6-agent Medallion architecture. Utilizes Foundation Model API embeddings, Vector Search for semantic nearest-neighbor evaluation, and automated Databricks Jobs API workflows. Refactored storage paths to native Spark Delta MERGE operations to guarantee transaction idempotency. | Databricks · PySpark · Delta Lake · Unity Catalog · MLflow · Vector Search · Next.js |
| ⚡ Velo: LLM Inference Gateway | High-performance, self-hostable, OpenAI-compatible LLM gateway written in Go. Implements semantic prompt caching with pgvector, dynamic request micro-batching, multi-backend failover routing, and Redis-backed rate limiting. Instrumenting the pipeline with Prometheus and Grafana demonstrated 130% higher requests per second and an 84% reduction in p50 latency at 60% prompt reuse. | Go · pgvector · Redis · Prometheus · Grafana · Docker |
| 🚨 HazAlert: Multilingual AI Emergency Dashboard | 1st Place out of competitive team entries at the Commit the Change x Google Gemini Prompt-a-Thon. Orchestrated 11 core Google Cloud and Gemini AI services into a cohesive voice-first dashboard. Developed low-latency telemetry translation pipelines and highly accessible frontend map routing workflows to assist users in identifying safer evacuation paths during crisis events. | Gemini API · GCP · React · Google Maps APIs · Core Translation |
| 🧠 Brian: Enterprise Agent Memory | Developed at LA Hacks 2026. Architected a stateful multi-agent system using LangGraph configured with custom Reader and Writer agents for multi-step execution over corporate data streams. Engineered a custom Model Context Protocol (MCP) tool server alongside a two-stage hybrid vector + BM25 RAG pipeline utilizing pgvector and cross-encoder reranking layers. | LangGraph · FastAPI · pgvector · MCP · TypeScript · PostgreSQL |
| 🚁 UAV Forge Perception Platform | Built automated synthetic-data pipelines in Python and fine-tuned deep learning YOLO architectures in PyTorch, increasing aerial target detection accuracy by 32%. Optimized edge deployment constraints by implementing INT8 post-training model quantization, reducing inference latency by 40% to meet low-power flight hardware thresholds. | PyTorch · OpenCV · YOLO · Vision Transformers (ViT) |
| 🛰️ UCI CanSat Telemetry & Control | Programmed modular, bare-metal flight software in C/C++ running on STM32 microcontrollers to support real-time sensor fusion and non-volatile flash-backed fault recovery. Developed an asynchronous Python ground telemetry application using PyQt5 and Matplotlib to parse and chart downlinked flight telemetry in real-time. | C/C++ · Python · STM32 · Embedded Systems · PyQt5 |
