Full-stack engineer · developer tooling & platform engineering
I build the tools other engineers work in — API testing platforms, engineering analytics, task intelligence.
_{Competitive programmer · olympiad mathematician · Baku, Azerbaijan}

• What I do: design and ship full products end to end — desktop clients, REST backends, admin panels, CLIs, CI runners and the deployment story around them.
• Where I go deep: multi-repo TypeScript platforms, test automation engines, on-prem/air-gapped delivery for banking-grade environments, and AI-assisted developer tooling (MCP servers, agent integrations).
• Belief: the design of a project matters more than its code — architecture that stays legible and a UI people actually enjoy beat clever implementations every time.
• Background: mathematics olympiads and ICPC-style contests — where I learned that the hard part is always the model, not the syntax.

A Postman-alternative built for teams that cannot use the cloud. Collections, environments, multi-step test scenarios, contract checks, system-design diagrams — in a desktop app, a CLI, a REST backend and an MCP server, deployable fully on-premise.

Eight repositories, one engine. callman-core is the shared request/scenario runtime — the desktop app, the backend, the CLI and the MCP server all execute requests through exactly the same code path, so a scenario behaves identically on a laptop, in CI and on a bank's air-gapped server.

Engineering highlights

• Protocol breadth beyond HTTP — Kafka produce/consume with a pooled client layer, PostgreSQL stored-function testing with transactional rollback, Redis and database steps inside a single scenario graph.
• Security & enterprise auth — OAuth2 flows, mutual TLS, LDAP/AD integration, Windows DPAPI-backed secret storage, scoped personal access tokens.
• Scenario engine — a directed graph of request / condition / loop / sub-scenario / AI nodes with variable propagation, retries, approvals and replayable run reports.
• AI, centralised — one prompt/parser layer in callman-core powering the in-app assistant, scenario generation and the MCP tool surface, with BYOK and managed key modes.
• Ships like enterprise software — Helm charts, offline install bundles, per-company desktop builds with electron-updater feeds, /ops JSON + Prometheus endpoints, in-app release notes gated in CI.

Multi-provider DORA, cycle-time, throughput and review-quality analytics for banking-grade engineering organisations. Self-hosted bundle or SaaS.

Ingests GitLab, GitHub and Bitbucket into one canonical data model, so the metrics engine and the UI never touch provider-specific types — a new provider is just another adapter.

NestJS · React + Vite · BullMQ workers · MongoDB · Redis · Turborepo · pluggable AI providers

Provider-agnostic task and delivery intelligence, built as a clean pnpm + Turborepo monorepo with its own MCP app so agents can work the backlog alongside humans.

NestJS (OpenAPI) · React + Vite · Tailwind v4 design tokens · shared TS/ESLint/Prettier presets · MCP

• Design first. Information architecture and interaction design before implementation — for the UI and the module graph.
• One engine, many surfaces. Shared runtimes over duplicated logic; if two clients can disagree about behaviour, they eventually will.
• Built for the worst environment. Air-gapped, proxied, LDAP-only, no internet — if it works there, it works anywhere.
• Automate the boring part. Codegen, CI gates, release tooling and AI agents wired into the actual workflow, not bolted on.

Languages

Frontend

Backend

Data & messaging

Infra & delivery

AI & tooling

_{Also: Arduino & embedded tinkering, Kali Linux, and far too much time spent on developer ergonomics.}

_{Currently building Callman — and shipping it into places the internet cannot reach.}