Senior Full-Stack Developer — Deploy & Extend CRM App (Telegram-first, Grok AI Integration)

Russian or Ukrainian speaking preferred for smooth communication. Hi I'm looking for a senior-level, highly experienced full-stack/backend developer to take a project from working prototype to a fully deployed, production-ready product — handling everything end-to-end with minimal back-and-forth. This is not an entry-level task; I need someone who can work independently, anticipate issues, and deliver a reliable, long-term stable result rather than a quick patch job. About the product: A web-based CRM primarily focused on managing conversations through Telegram (WhatsApp support is a secondary/optional add-on - see below), from a single dashboard. Incoming messages are processed by an AI layer using Grok's API, with different pre-set moods/styles selectable per conversation (Grok's built-in mood/personality presets make responses feel more natural and varied compared to a single fixed tone). Replies are sent with realistic human-like delays (not instant). Everything is monitored and controlled through a web panel. Core scope (required): — Review and extend an existing codebase (I'll share via GitHub/zip) includes a Telegram module (via the telegram/GramJS Node.js library, connecting to an existing account rather than a bot), basic AI logic currently wired to Claude API, and a web dashboard skeleton — Integrate Grok's API as the main AI layer, including support for selecting between its mood/style presets per conversation — Solid, production-grade Telegram integration — this is the primary channel and needs to be rock-solid (reliable reconnects, no dropped sessions, clean handling of rate limits) — Dashboard: conversation list, full message history, per-conversation "AI auto-reply / manual mode" toggle, manual message input, mood/style + system prompt settings panel, delay/typing-speed configuration — Additional feature — client info panel + escalation alerts: A small side panel next to each chat showing key notes/facts the contact has shared during the conversation (e.g., name, preferences, important details mentioned) — manually visible for quick reference, and also fed back into the AI's context so it remembers this information in future replies The AI should be able to flag when it's unsure how to respond or hits a difficult/sensitive question, and notify me (e.g., via a dashboard alert or notification) that manual intervention is needed Specifically: if a contact directly asks something like: "are you an AI?" / "am I talking to a bot?", the system should flag this conversation for manual takeover rather than letting the AI improvise a response — Set up deployment (GitHub repo + hosting — Railway or Render, your recommendation welcome), configure environment variables, database, dependencies — Deploy to a live, publicly accessible URL running 24/7 — Walk me through one-time Telegram authorization (SMS code) — Full end-to-end testing — Hand over the live link plus full admin access (hosting/repo) — Clear documentation (Russian is fine if easier) Optional add-on scope (quote separately): WhatsApp integration (e.g., via Baileys), connecting to an existing WhatsApp number via QR code, added as a second column in the dashboard alongside Telegram. Open question for candidates — tech stack: The existing prototype is in Node.js, but I'm open to switching the core to Python if you believe it results in a more stable, maintainable long-term product — for example, using Telethon instead of GramJS for the Telegram layer. Please share your recommendation and reasoning with your proposal; I care more about long-term reliability than sticking to the current stack. What I'm looking for in a candidate: — Strong, verifiable track record with backend development, API integrations, and deployment (please share relevant past projects/portfolio) — Experience with LLM API integrations (Grok, OpenAI, Anthropic, or similar) — Experience with Telegram automation (bots or userbots) — this is the most important technical skill for this job — Comfortable owning a project end-to-end without heavy hand-holding, and giving honest technical recommendations