Hi, I'm Al
CS senior at The City College of New York, graduating December 2026.
I work across full-stack web development, agentic AI, and machine learning — with a growing focus on information security. Recently shipped five production security dashboards and cloud data pipelines at Ariat, a full-stack vendor-finder app, and a multi-agent reinforcement learning system for connected self-driving cars.
I started building software during a web-development internship at The Difference App, a digital health company in New York. I updated their site, improved SEO using tools like Yoast and SEMrush, built Power BI dashboards to summarize weekly traffic, and wrote PowerShell and Excel/VBA macros to automate repetitive reporting.
Since then I've been building my own tools on GitHub. The one I'm proudest of is an App Health Dashboard — a Flask + Chart.js view of error rates, open bugs, and failing endpoints — that grew directly out of the dashboards I built at my internship.
I'm also leading CCNY senior-project research on multi-agent reinforcement learning for connected self-driving-car routing, using JAX with the Waymax and AlphaSim simulators. And this summer I worked as an Information Security Engineering Intern at Ariat International, where I designed and shipped five production security dashboards and the cloud data pipelines behind them — combining the dev work and data analysis I've been doing with a real security context.
At a glance
Full-stack · Observability
One page instead of three log tabs.
A Python + Flask web app that ingests error logs, a bug list, and support tickets from CSV/JSON feeds and surfaces the metrics that matter on a single page — errors in the last 24 hours, open bugs by severity, top failing endpoints, and average uptime. Includes a Power BI export so the team can slice the same data without re-querying. Built during my web-dev internship at The Difference App.
Full-stack · Team project
Find nearby street vendors and see what they're selling.
A responsive full-stack web app that helps people locate nearby street vendors and browse their menus. Vendors register, post their location and hours, and publish a menu; customers search by location or food type and view vendor details on a map. Built as a four-person team project for CCNY's Software Engineering course (CSc 47300) — a React + Vite single-page front end wired to a Supabase (PostgreSQL) backend for auth, data, and real-time updates.
Automation · Log analysis
Triage tool that reads three log shapes and reports what's wrong.
Python and PowerShell automation that reads three log formats (Python app logs, web-server access logs, structured NDJSON) and flags patterns that usually mean something is wrong: a burst of errors, a repeated exception class, errors right after a deploy, slow endpoints. Each finding is scored and written into a Markdown report sorted by severity so I can read it with morning coffee.
AI/ML · Reinforcement learning
Q-learning and a from-scratch Double-DQN, written in pure NumPy.
A reinforcement-learning sandbox with two agents I wrote by hand: a tabular Q-learning agent and a Double-DQN with a 2-layer MLP (replay buffer, target network, gradient-norm clipping). Trained on GridWorld and CartPole. I wrote the neural network's forward and backward pass by hand to actually understand the math behind the RL papers I read for my CCNY research on self-driving-car routing.
AI Research · Multi-agent RL
Teaching a fleet of self-driving cars to route packets as a team.
My two-semester CCNY senior-project research (CSc 59866). Instead of one central controller, eight connected autonomous vehicles each run their own reinforcement-learning policy and cooperate to route packets through a vehicle-to-vehicle wireless network — even under up to 25% link loss. I built and compared three algorithms (MARL-PPO, MADDPG, and MAPPO with agent-to-agent messaging) against shortest-path and greedy baselines on a JAX backend with the Waymax and AlphaSim simulators; all RL methods held the latency-violation rate at 0% across every loss level.
Cryptography · Course project
Encrypt, decrypt, and crack Vigenère ciphers.
A Python + Flask web UI and CLI for polyalphabetic substitution ciphers, plus a cracker that recovers the keyword from ciphertext alone using Kasiski examination and Index of Coincidence. Built for the group project in CCNY's Computer Security course (CSc 47200) — my foundation in applied cryptography.
Updated site layout and content, improved SEO (Yoast, Sitechecker, SEMrush), built Microsoft Excel and Power BI dashboards summarizing weekly traffic and content engagement, and wrote PowerShell + Excel/VBA macros to automate the team's repetitive reporting.
Paid, competitive, hands-on program building real-world mission-driven applications alongside software professionals. Learned user-centered design and prototyping, designed algorithms for application use cases, and built interactive websites using HTML5, CSS, and JavaScript.
Mentored teens through STEM courses, including basic concepts of machine learning, AI, and model evaluation; coding games with Java; and teen coding with Python.
Languages
Frameworks & libraries
Tools & platforms
Concepts