I build cloud-ready systems, DevOps pipelines, AI platforms, and research infrastructure. My work usually sits where software engineering meets automation, data, and production-style thinking.
Cloud / DevOps -> Docker, Kubernetes, Terraform, CI/CD, monitoring, automation
AI Platforms -> RAG, Graph-RAG, LLM agents, vector search, evaluation pipelines
Research Systems -> fuzzing, Bayesian risk estimation, reproducible experiments
Backend / Data -> Python, SQL, APIs, ETL, validation, dashboards, workflow state
|
Building repeatable infrastructure, deployment workflows, testing pipelines, observability, and automation-heavy systems. |
Building RAG systems, metadata pipelines, vector retrieval, LLM evaluation, and evidence-backed AI workflows. |
|
Designing reproducible experiments for fuzzing, software testing, Bayesian risk estimation, and local LLM robustness. |
APIs, persistent logs, workflow state, source-grounded outputs, dashboards, validation checks, and measurable reliability. |
|
DeltaRAG + Graph-RAG system that monitors Android ecosystem changes and generates evidence-backed risk tickets.
|
Reproducible fuzzing infrastructure with Bayesian estimators for measuring residual software risk.
|
|
Document intelligence platform for large PDF collections with metadata-enriched RAG and source-grounded answers.
|
Recommendation service with model training, API serving, Docker, CI/CD, monitoring, drift checks, and A/B testing.
|
| Area | Project | Why it matters |
|---|---|---|
| Risk Intelligence | ECI Pipeline | DeltaRAG + Graph-RAG pipeline that turns Android ecosystem changes into evidence-backed risk tickets |
| Software Testing | FuzzBench Research Pipeline | Reproducible fuzzing infrastructure with Bayesian residual-risk estimation |
| AI Platform | MetARAG | RAG chatbot for large PDF collections with metadata enrichment and source-grounded retrieval |
| MLOps | Movie Recommendation | Model training, API serving, Docker, CI/CD, Prometheus, Grafana, drift checks, A/B testing |
| Security | MTProto 2.0 | End-to-end encrypted messaging protocol implementation |
| Reasoning | Chain-of-Thought on CLEVR | Evaluating supervised reasoning traces for visual question answering |
flowchart LR
A[Bayesian Residual Risk of Fuzzing] --> B[FuzzBench + Docker + LLVM]
C[Counterfactual Fact Verification] --> D[FEVER + Phi-3 Mini + Robustness]
E[Chain-of-Thought on CLEVR] --> F[VQA + Reasoning Traces]
-
A Bayesian Approach to Estimating Residual Risk of Fuzzing
Building statistical estimators and reproducible fuzzing workflows for measuring residual software risk. -
Counterfactual Fact Verification
Evaluating local LLM robustness on FEVER using structurally tiered claims and counterfactual variants. -
Chain-of-Thought on CLEVR
Studying how supervised reasoning traces affect visual question answering behavior.
| Role | Organization | Focus |
|---|---|---|
| Graduate Research Assistant - DevOps Engineer | University of Illinois Chicago | Docker, FuzzBench, LLVM, reproducible testing pipelines |
| Graduate Research Assistant - AI Platform Engineer | University of Illinois Chicago | RAG, LLM evaluation, document intelligence, metadata pipelines |
| Trainee Software Engineer | Mu Sigma | Python, SQL, ETL automation, data quality, analytics workflows |
I am open to roles and conversations around DevOps, Cloud Engineering, Platform Engineering, AI Infrastructure, Backend Engineering, and Applied AI systems.
