// Hire Me
Reliable AI infrastructure, shipped
As a software engineer specializing in cloud architecture and AI/ML systems, I help organizations turn raw enterprise data into intelligent, production-grade products. From high-throughput ingestion pipelines processing 10k+ documents a day to GenAI-driven MLOps that cut release cycles by 35%, I build the infrastructure that makes AI dependable at scale — and I share what I learn by mentoring engineers and publishing hands-on resources like the ones in this portfolio.
Written for AI/ML hiring managers and technical recruiters evaluating my work, engineering collaborators and mentees, and fellow AI/ML practitioners and learners.
// 01 · What I Do
What I do
RAG & ingestion infrastructure at scale
At Amazon, on the AWS Quick RAG team, I build high-throughput ingestion pipelines for Amazon Quick Suite, Q Business, and Kendra that process 10,000+ documents a day and power RAG and LLM experiences over enterprise data — including index re-ingestion workflows that eliminate manual operations and enable zero-downtime updates as models evolve.
MLOps & CI/CD for ML
I streamline index ingestion by integrating validation and safe rollout mechanisms into CI/CD workflows for ML teams, delivering 60% faster deployments. Earlier, as a Cloud Engineer at People Tech Group, I led a GenAI-driven MLOps solution that turned natural-language prompts into validated Infrastructure-as-Code, cutting release timelines 35%.
Applied ML end-to-end
Haven, my local AI therapy companion, is a full walk of the ML lifecycle on an 8GB laptop GPU: dataset cleaning, EDA, feature engineering, a QLoRA fine-tune of Llama 3.2 3B, a trained 28-emotion classifier, and honest evaluation — validation perplexity dropped from 32.6 to 9.79.
Responsible-AI system design
When Haven's fine-tune regressed on crisis safety, I moved the guarantee out of the model into a server-side, regex-verified safety layer no future training run can trade away. Triage Copilot, an on-call AI agent, is built the same way: it advises but never executes, and states a confidence level and a falsification condition for every conclusion.
// 02 · Selected Work
Selected work
The Applied AI Brief — A Published Industry Column
A published AI industry column with three issues live — plus the double-opt-in email platform behind it: Lambda, DynamoDB, and SES on a dedicated AWS account, deployed by OIDC with no stored credentials.
View artifactHaven — Anatomy of a Listening Machine
A fully local AI therapy companion — QLoRA fine-tuned Llama, a trained 28-emotion classifier, retrieval-grounded answers, and hands-free voice — every step of the ML lifecycle on an 8 GB laptop GPU. Interactive build log with demo video on its page.
View artifactTriage Copilot — On-Call AI Assistant
An AI agent that advises but never executes — it questions its way to a hypothesis for a failed AWS pipeline and refuses to guess. Built for a 3am page. Live demo on its page.
View artifact// 03 · Credentials
Credentials
MS, Cyber Security & Privacy — New Jersey Institute of Technology (NJIT)
Graduate studies, AI & Machine Learning — Indiana Wesleyan University (in progress)
AWS Certified Solutions Architect – Associate
AWS Certified Developer – Associate
CompTIA Security+
Certified Ethical Hacker (CEH) v12 — EC-Council