From shipping scalable AI systems, cloud‑native infra, and seriously fast CI/CD, I build production‑grade ML/AI
applications and AWS‑first architectures.
Strong in observability, automation, and cost‑efficient design.
LLM inference on AWS Lambda (C binaries / llama.cpp)
45+ models
Unified routing & access via Python lib
15× faster
28 Docker images: 30m → 2m via parallel builds
Recent Work
ML Ops & AI Development Engineer
Enterprise Technology — Tempe, AZ
–
LLM & AI Application Development:
Engineered a Python library facilitating access to over 45 Large Language Models
(LLMs), with routing across multi-modal pipelines using AWS Lambda and API Gateway.
Deployed an LLM inference solution on AWS Lambda using llama.cpp C binaries, achieving
10 tokens/sec.
Built a multi-threaded Python app for Raspberry Pi integrating text-to-speech, speech-to-text, facial
recognition, and a local LLM.
Fine-tuned LLMs and created proxy servers for a Model-as-a-Service framework to host and route
open-source models.
Implemented rate-limiting and caching with Lua scripting on ElastiCache Redis using sliding windows.
Cloud Infrastructure & DevOps:
Provisioned scalable, multi-stage AWS infrastructure using Terraform.
Architected CI/CD pipelines with parallel Docker builds on Jenkins + Kubernetes, cutting build time
for 28 images from 30m to 2m.
Designed microservices using AWS Lambda, DynamoDB, S3, API Gateway; integrated OAuth2 with Google
Drive.
Implemented observability via SNS alerts; reduced downtime and improved traceability.
Automated secure cross-cloud provisioning across AWS and GCP.
Integrated Redshift and DynamoDB using zero‑ETL to build analytics-ready datamarts.
Developed Firecrawl, a caching layer for web crawling to reduce external API usage and cost.
Graduate Student Researcher — Thesis
Bio‑Inspired Robotics, Technology & Healthcare Lab — Tempe, AZ
ROS2 PID control for UR‑16e robotic arm enabling adaptive load‑carrying tasks.
ROS‑based SpaceMouse controller for real‑time manipulation.
Selected Projects
Papertrail — Multi‑Modal OCR Document Organizer
Python • OCR • AWS • OAuth
Document organizer with fine‑grained access control & sharing. Multi‑modal OCR endpoint for accurate
file→text conversion. Google OAuth2 (Drive & Picker) integration.
Decentralized quadcopter swarm with Potential‑Field & Frontier Exploration for dynamic 3D mapping;
validated 100×100 grid mapping in Gazebo while avoiding local minima.