
Telegram → SendPulse Automation
Turned manual Telegram contact transfer into an always-on email acquisition and synchronization pipeline.
Backend systems, automation pipelines, document AI, API integrations and recovery of difficult legacy projects — from technical audit to deployment and operational handoff.
Each case focuses on the engineering problem, the decisions made and the delivered result — not just a list of technologies.

Turned manual Telegram contact transfer into an always-on email acquisition and synchronization pipeline.
Production backend work around async orchestration, transactional consistency, external AI/data services and failure recovery.
$ git deploy dev ✓ nginx / PHP-FPM / MariaDB ✓ GitHub release workflow ✓ admin restored ! legacy HTTP 500 traced ! malicious module code neutralized ✓ reproducible DEV environment
Built a reproducible DEV environment, replaced manual SFTP deployment and stabilized critical failures in a damaged legacy stack.

Prepared a SaaS integration for renewed Google review after a previous rejection, including OAuth flow, demo and compliance materials.
Multi-modal extraction pipeline for invoices and sewer inspection reports across text PDFs, scans, phone photos and handwritten forms, with a reviewer screen for operator correction and extraction-quality signals recorded for later confidence-based routing.
Recovered the iOS build path, deployed the backend to Azure, validated Work Orders and isolated production crash/offline-sync failures.
FastAPI, REST, PostgreSQL, SQLAlchemy, Redis, asynchronous services, migrations and integration boundaries.
Telegram, external APIs, background processing, retries, scheduling, synchronization, logging and operational handoff.
Text extraction, OCR/vision fallback, structured extraction, classification, validation and multi-modal pipelines.
Legacy codebases, build environments, production failures, deployment paths, root-cause analysis and stabilization plans.
I am most useful on projects where the difficult part is not writing one more endpoint — it is understanding where the system actually fails, choosing a maintainable boundary and getting the whole path to work under real conditions.
That can mean tracing a broken legacy mobile build, validating an OAuth approval path, making a data pipeline resilient to external failures or separating environment problems from business-logic defects.
Send a short description of the problem and the current state of the system — I’ll reply with a first assessment and next steps.