MSCS @ Northeastern University (Boston) • I build AI-powered apps and the platforms that keep them fast and reliable.

• Built SMART AI Chat, a public-facing LLM 311 assistant for the City of Fort Wayne, on LangChain + the OpenAI API
• Cut cold-start latency 100× (20s → 200ms) with Redis pipelining, and page load 5× (30s → 7s) with Brotli/Gzip and a 60% smaller bundle
• Reproducible CI/CD with Docker multi-stage builds on GitLab CI across 40+ services
• Rust-first tooling for deployment automation and developer workflows

Looking for: summer 2027 internship / co-op roles in AI/ML or platform engineering.

Links: Blog • Resume • LinkedIn • Email

• Languages: Rust, Python, TypeScript/JavaScript, Java, SQL, Bash
• App/AI: LangChain, LangGraph, Google ADK, MCP, FastAPI, Flask, Spring Boot, Angular, React, Postgres, Redis, WebSockets
• Infra/DevOps: Linux, Docker, GitLab CI, Nginx, Cloudflare, OpenTelemetry, Prometheus/Grafana, Kubernetes, NixOS

• Hikari — GitOps platform in async Rust, a lightweight Kubernetes alternative for small VM fleets Rust + Tokio + Axum + Postgres/SQLx • WebSocket agent/server design for real-time config propagation Each agent reconciles its host against the last-applied declarative JSON config; AES-256-CBC config secrets under an RSA-4096 OAEP envelope. Nine tagged releases. Write-ups: Agent/server design • Daemon mode

• Patternsmith — multi-agent LLM pipeline that turns a topic into a publication-ready tutorial Python + LangGraph + Pydantic v2 + Hugo • Research/Teaching → Code → Writing → validation over one typed state Generated code is embedded verbatim, so it never drifts from the prose or the Mermaid diagrams. Model-agnostic behind a ChatLLM protocol: the same pipeline runs against a hosted API or a local model (Ollama / LM Studio) on one config flag. MIT licensed. Site: patternsmith.distroy.dev

• Documan — a college digital library where every subject has an address: department, year, semester Ran in production on Flask/Postgres for five years (Jun 2020 – May 2025): 4,000+ students, 200,000+ file deliveries, 1,200+ active users, 99%+ uptime, over a web app and a Telegram bot. Then rewrote it from scratch as a Java 25 / Spring Boot 4.1 API (virtual threads, Redis caching, Cloudflare R2, Meilisearch kept in step with Postgres by a transactional outbox) plus an Angular 21 signals client, identity on Clerk. The rewrite has no users; it is where the interesting engineering is. Write-up: Zero-cost digital library

• Autodeploy — single Rust binary that turns a TOML config into a working local deployment Interactive CLI (git2, inquire) that generates Docker Compose at runtime for multi-service stacks. Cut local environment setup from an estimated 2 hours to 20 minutes. Write-up: Hassle-free deployments

• LangChain #19427 — fixed a Pydantic serialization bug in the GPTCache integration, where a ChatGeneration returned by ChatOpenAI was not handled on cache lookup. Merged Mar 2024.

• Best Podman Container Implementation and People's Choice — Pods, Prompts & Prototypes Hackathon (Red Hat / IBM), Mar 2026, for OAAM (Open Agent Applet Marketplace): a local-first AI web app that extracts structured data from documents fully on-device via RamaLama + Podman AI Lab.
• Star of the Month — Nebulogic Technologies, May 2024.

• Observability for modern workloads: metrics, logs, tracing
• DNS privacy and running your own resolver stack
• Deployment automation lessons from Hikari and Autodeploy

Read here: Blog home