Software that fits your workflow.

Custom web and AI-powered apps, shipped end-to-end.

From domain and DNS through backend, frontend, and LLM pipelines — delivered on modern cloud, then maintained.

Capabilities

Claude-powered content pipelines, structured outputs, semantic clustering, retrieval, voice-aware generation.

Python · FastAPI · Flask · SQLite · SvelteKit · Tauri · PWAs for web, mobile, and desktop.

Railway deployment, Cloudflare CDN and security, CI/CD, monitoring, automated backups.

Domain setup and transfers, DNS, DKIM/SPF/DMARC, verified sender domains, transactional email.

Fixed-price builds with a scoping call up front. Ongoing maintenance available on every engagement.

🚀

Starter App

From $5,000

2–3 weeks

Good for a focused client portal or internal tool.

aupaatleti.com ↗

Problem: Following Atlético Madrid means juggling Spanish sports press, club channels, Twitter, and Reddit — most of it not in English. Getting genuinely caught up took an hour every morning.
Built: A daily intelligence briefing that does the catch-up for you. Pulls from 21 Twitter/X accounts, 14 RSS and YouTube feeds, and Reddit — translates Spanish and Portuguese sources automatically, classifies each story by injury, transfer, lineup, match, or tactics, and surfaces the top signals each morning by email and push notification. Stories confirmed by multiple independent sources rank higher. Fan sentiment is tracked per post and aggregated into a daily mood indicator with a 30-day trend.
Admin: Source health grading (A–F), keyword management with usefulness scoring, one-click AI audit recommendations, pipeline diagnostics, and an AI-generated daily tweet in selectable character voices (Simeone, Griezmann, Costa) for distribution on X.
Stack: Python · FastAPI · SQLite · Claude Haiku + Sonnet · Resend · Railway · Cloudflare.

local-first · mac app

Problem: Personal finance tools want your bank credentials and a monthly subscription. There was no serious option for people who want real financial depth — retirement projections, tax estimates, debt amortization — without handing their data to anyone.
Built: A local-first Mac desktop app. Net worth dashboard, investment portfolio tracker, FIRE retirement calculator with Monte Carlo stress testing, 2025 federal tax estimator, debt amortization, college and 529 projections, and Bitcoin/ETF tracking. All data lives in a SQLite file on your machine — no Plaid, no cloud sync, no subscription, no licence gate.
Context: A self-built product distributed via Gumroad — not a client engagement, but a live example of the full-stack and local-first desktop capability on offer.
Stack: Tauri 2 · SvelteKit · TypeScript · SQLite · Vitest.

Problem: Lisa runs online fitness coaching. She needed her clients logging check-ins and tracking progress in a branded environment — not a generic SaaS that puts its own logo on her product and competes with her on price.
Built: A fully branded client portal. Participants join 30-day bootcamps, log daily habit check-ins (cardio, water intake, sleep, mindfulness), set weekly goals, and compete on a leaderboard. Lisa manages bootcamps, reviews per-participant progress, and sends a custom newsletter — all under her own brand and domain. Clients see The Fitish, not "whatever platform."
Stack: Python · Flask · SQLite · Plotly · Resend · Railway · PWA.

redgymreps.com ↗

Problem: A group of D1 college athletes heading into summer internships needed to maintain structured training away from their teams — with a coach who could monitor, adjust, and review their progress remotely. Nothing on the market supported that loop.
Built: A mobile-first training tracker for remote athlete-coach pairs. Each session is pre-loaded from the athlete's 12-week periodized program, with target weights calculated from their 1RMs. After each session, progression hints (add weight / hold / back off) are generated from the logged RPE. Three built-in programs: a strength block for competitive swimmers, a general fitness recomposition plan, and a sub-2:00 half marathon plan — each running 3–4 phases of linear periodization. Athletes also log a secondary activity (swim with race times, daily steps, or pickleball), all included in an end-of-program report covering PRs, volume, adherence, body weight trend, and coach feedback.
Coach view: Full plan editing mid-block, session comments per athlete, adherence dashboards across the full roster, and a complete audit log of every plan change.
Stack: Python · Flask · SQLite · Plotly · Gunicorn · Railway · PWA.