App-Kit
Multi-tenant SaaS platform: build Telegram bots, mini-apps and sites from pluggable modules.
Projects
Each case study below follows the same structure: the problem, what I built, the engineering details and the result. Numbers come from the repositories, test suites and real runs; where a figure is an estimate, the case says so.
App-Kit is a multi-tenant platform for sites, Telegram bots and mini-apps, with 15 pluggable modules and payment, booking and notification providers behind adapters. Lingvo puts realtime speech translation and AI meeting notes on one Django backend, with a GPU agent on the audio path. seo-audit crawls a site, runs 400+ checks and turns the facts into a weighted score and a PDF report.
The cases share a few habits. Architecture decisions are written down as ADRs (31 in App-Kit, 13 in Lingvo), external services sit behind adapters, and heavy or risky work such as GPU inference, headless browsers and crawling arbitrary URLs runs in separate queues or services, away from the request path.
The common core is Python, Django, DRF, Celery, PostgreSQL and Redis, with Docker for delivery and Next.js where a product needs its own frontend. Commit activity next to each case comes live from my self-hosted Forgejo. For work history before these products, see the CV.
Multi-tenant SaaS platform: build Telegram bots, mini-apps and sites from pluggable modules.
Two products on one backend: realtime speech translation and AI meeting notes.
Automated SEO, technical SEO, performance and AI-readiness auditor in Python/Django: a URL goes in, 400+ checks, a weighted score and a PDF report come out.