Full SaaS trading education platform built solo from zero to production. It supports real users and paid Stripe subscriptions, with paper trading, live market data, AI-assisted journaling via OpenAI, 24/7 automated monitoring via Puppeteer, real-time charting, web push notifications, and transactional email. I designed and built the full backend, including auth, billing, cron jobs, and storage.
Enterprise-grade dealership management system built as a white-label SaaS. Full CRM with lead tracking, inventory management, appointment scheduling, financial reporting with interactive charts, and admin/sales role separation. Relational data model using Prisma ORM over PostgreSQL, a deliberate choice to work with structured data at scale.
Precision calculator for perpetual futures traders. Handles weighted average entry price across chained position adjustments, direction-aware PNL for long and short positions, leverage-aware notional sizing, live liquidation price, and risk/reward ratio. Built entirely in TypeScript with Jest test coverage. The math is exact, not approximate.
Full e-commerce platform built for a clothing brand from scratch. Product catalog, shopping cart, Stripe checkout, order management, and AWS S3 for media. My first major project, built in Django/Python after bootcamp in 2020 and relaunched with improvements. The foundation that started everything.
iOS app for cleaning up cluttered photo libraries with a Tinder-style UX. Swipe left to delete, swipe right to keep. Drag photos into album nodes for organization. My first React Native / Expo app shipped to the Apple ecosystem. It taught me mobile-specific patterns including gesture handling, native permissions, and device storage access.
Open-source humanitarian platform connecting amputees in conflict zones with volunteers willing to 3D-print prosthetics. AI-powered content moderation via OpenAI and NudeNet, member verification with a 48-hour review pipeline, and Stripe donation processing. Built in Django on Railway. It is the project I'm most proud of for reasons beyond code.
PWA route tracker built to solve a real problem: the company app for NEMT drivers couldn't export any data. Logs daily routes, stops, mileage, and patient trip info with an interactive Leaflet map, OpenRouteService geocoding, and offline-capable localStorage persistence. Built out of necessity, used in the real world.
Interactive bilingual family registry mapping 142 members across graph, 3D, and olive-tree views. It includes generation-aware layouts, search, member profiles, photo management, role-based access, statistics, and responsive navigation for a large connected dataset.
Browser-based music experience that turns keyboard input into sound and animated 3D visuals inside a navigable space scene. It includes recording and playback, camera and visual modes, mobile controls, performance presets, and Firebase analytics.
Private agent systems built for real workflows, combining messaging, browser automation, market data, memory, and guarded execution.
Built ChartSensei as a private second developer dedicated to ChartSense, the project featured above. It does not start from a blank chat. It carries the architecture, deployment setup, file-level knowledge, prior decisions, known bugs, and roadmap across sessions, so I can send a Telegram message or voice note and continue where the work actually stopped.
ChartSensei works directly inside the main application and its separate 24/7 trade monitor. It reads and edits code, traces production issues, runs builds and tests, checks repository state, and prepares Railway releases. Significant work is written back into durable project memory, while credentials, Stripe changes, destructive commands, and Git pushes remain behind explicit boundaries.
Inside Telegram, I split ChartSensei into dedicated topics for General, Dev / Deploy / Debug, Video / Content / Marketing, and SEO & Analytics. Each topic has its own operating context while sharing the same long-term ChartSense memory, so it can move from a production incident to a release, content planning, or live GA4 and Search Console analysis without losing the thread. As the one person behind ChartSense, that structure makes ChartSensei my only day-to-day teammate, carrying work that would normally be divided across engineering, operations, growth, and analytics.
Built Alfred as my private personal butler: a persistent agent with one shared memory across Telegram, voice calls, and workspace files. Alfred is designed to know my context, help with research and development work, challenge weak thinking, remember what matters, and use proactive heartbeats for email, calendar, weather, and other routine checks without becoming noisy.
The current version is the digital foundation for a future physical assistant. Its voice layer connects Twilio, ElevenLabs, and a FastAPI bridge so phone conversations use the same memory as text sessions, then write post-call summaries back into the workspace. As robotics tools mature, Alfred is intended to be the first agent I move from screen and phone into an embodied robot assistant.
Built Dexter / ChartDoc as one private trading research brain around a proprietary TradingView Pine strategy. The strategy runs on TradingView and emits structured alerts; Dexter reads those alerts with analyzed trade ideas, source and news context, cross-asset pressure, derivatives behavior, open exposure, and runtime health before deciding whether a paper trade should be ignored, watched, reduced, protected, or submitted inside ChartSense.
The important part is the feedback loop, not the signal recipe. Every candidate, missed setup, fill, exit, source reaction, and review becomes evidence. Over months of live collection, 1,600+ strategy-alert rows, 340+ analyzed setup records, source-to-market attribution, rule regression, shadow-policy testing, and post-trade forensics have trained the agent to recognize clean context, crowded late entries, degraded risk/reward, and moments where restraint is the better trade.
Built FarisWayne as the internal workflow agent behind Debt Clinic consultations, designed to absorb the tedious appointment workload that normally gets scattered across inboxes, CRM tabs, text threads, and calendar checks. It watches Acuity and Gmail, separates routine bookings from exceptions, and routes each case into the right Telegram topic: new appointments, cancellations, low-debt reviews, CRM duplicate warnings, invalid or international numbers, client replies, and manual follow-ups. It also runs the SMS lifecycle around every appointment: booking confirmation, 6 AM check-in, 2-hour reminder, 30-minute reminder, reply detection, and suppression logic so responsive clients are not spammed.