Dylan Hans — Projects

architecting a secure, VM-based AI agent platform that enables autonomous execution of end-to-end knowledge work across the CLI (PROTECTED), enterprise tools, codebases, browsers, and cloud deployments. NexusOS empowers organizations to offload repetitive, low-value tasks to intelligent agents while maintaining strict security and system-level isolation.

with a voice-first interaction model powered by WebRTC streaming and WebSockets, NexusOS orchestrates LLM-driven pipelines that translate natural language into structured JSON actions for deterministic execution via MCP servers. Agents can perform operations across VSCode-like environments, productivity suites (Excel, Word, Gmail), and browser automations (Selenium), simulating human-like cursor-driven workflows while preserving backend efficiency and OpenAI Whisper text to Speech for real-time token stream and feedback.

the current tech stack -> Linux VMs via AWS EC2 servers, ECS, Node.js + Python hybrid services, PostgreSQL, and containerized deployments on AWS, GCP, and Azure. Cloudflare, Sentry, AWS BS. NexusOS integrates sandboxing, secure file/resource sharing, & persistent state management to safely run code, automate CLI commands, & deploy demos in isolated environments.

these capabilities will evolve into a production-ready infrastructure designed for enterprise pilots, with observability, CI/CD, TDD, and infrastructure-as-code to ensure resilience and scale.

at this stage for building a thin mvp, projected outcomes include 30–40% faster task completion, 2–3x worker productivity uplift, & potential cost of < ~$0.80 per 12 hours for average worker load with OpenAI’s GPT. For higher-level tasks, claude, gemini, and other models will be available.

In Development | Links: GitHub, Website, App

collection of Jupyter Notebook, markdown files, oneNote schematics, projects, labs, etc. with walkthroughs. These focus on machine learning and AI first approaches, ranging from LLM strategies to implementation.

Links: Github (Private), Collection

developing a cutting-edge banking platform that empowers users to finance in-store purchases by seamlessly integrating with their preferred digital wallets. Split Financial offers flexible, interest-free options for qualified users, along with the choice for long-term financing.

with swift and various frameworks, this application enables users to divide payments into manageable installments with minimal to no interest after a short application that performs a soft credit check and a quantitative model to evaluate user eligibility through calculating metrics from their retrieved equifax credit bureau data. The current tech stack is react, tailwind, next.js, typescript, redux, python, on AWS, integrating google cloud, plaid, dwolla, appwrite, and firebase auth.

these will change in the near future to the appropriate infrastructure that can handle banking institution loads and scalability. At this time in the building stage, users can sign-in through their existing banking institutions using Plaid for secure transactions and financial privacy.

Halted | Links: GitHub, Website, App

–

In Development | Links: Github, DashBoard

engineered a Python-based financial analysis assistant on Streamlit integrating OpenAI interaction capabilities with real-time stock price retrieval, technical analysis (SMA, EMA, RSI, MACD), financial reports, risk, media, and interactive plotting functionality.

Links: GitHub, Platform

designed several prediction models using Python and Streamlit that provide invaluable insights into option valuation, enabling investors and financial professionals to make informed decisions. These options give the holder the right to buy (call option) or sell (put option) an underlying asset, such as a stock, at a specific price on or before a predetermined expiration date.

Links: GitHub, Platform

educational platform that empowers teaching professionals’ to efficiently assess student learning objectives and streamline report card generation, through utilizing machine learning models and NLP to analyze student performance and deliver feedback.

leading end-to-end development of a dynamic, responsive web application using React, Vite, TailWind, Python, NodeJs, Redux, MongoDB and Google Cloud, etc. ensuring scalability, performance, robust functionality, and optimal user experience.

Links: GitHub, Website

led the team in software design, fabricated UI controls for the mechanical linkage system using C++, microcontrollers, Python with OpenCV, NumPy, SciKit, etc. for computer vision to detect human hand gestures, and manage item retrieval with force sensitive resistors FSR402.

implemented algorithms to direct pressure expenditure, reduced contraction duration to 0.8 seconds, handling 3kg+ payloads with a weight of 750g.

supervised under a Professor, the prototype is to be integrated onto the faculty owned Franka Emika Robotic Arm after placing in the top teams of the Capstone Engineering Competition and may receive further funding or research.

Links: GitHub, Course Page

built a compiler and language processor fo Quby, a modern language that integrated components of C, Ruby, Pascal, and Turing on Linux, leveraging IBM’s construction technique. Adapted from PT Pascal compiler using S/SL, project’s outcome & tests sets revealed an evaluation score of 97%.

Links: GitHub, Course Page

constructed a VR environment in Unity with a physical feedback model using microcontrollers and programming in C, simulating layers of soft tissue and bone as a real-time positional response system. Led the hackathon team in software development and earned 2^nd overall.

Links: GitHub, LinkedIn

directed the software engineering team in constructing a full-stack trial analysis website to inform consumers on real-time product inventory inflation or price adjustments in their local stores. Applied HTML, CSS, Docker, C++, Bootstrap, etc. ranked top in the SAC.

Links: GitHub, Course Page

designed CLI marketplace system for a software ‘company’ specified by a set of their requests. Assessed requirements, specifications, and met design quality plan.

application includes register functionality, login interface, product descriptions and account information through the use of Python, team Sprints, SQL, Docker, containers, and CI/CD. Measurement and software inspections conducted, integration tests were completed.

Links: GitHub, Course Page

created an autonomous vehicle in MATLAB & C to record data when analyzing potholes in a simulated real-world domain using lidar, radar, and camera modules. Integration analysis demonstrated savings of $100 million over 10 years for the City of Toronto.

Links: GitHub, Report