Full-stack Developer

Job Title: Full-stack Developer (Agentic AI & "Vibe" Software Engineer – FinTech/CPA AI)

Company: Loopnow Technologies Private Limited

Location: Work from office — Sonam Shopping Centre, Mira Bhayandar Road, Old Golden Nest Phase 2, Queens Park, Bhayandar East, Mira Bhayandar, Maharashtra, India (Targeting applicants from the Mumbai Metropolitan Region)

Job Type / Shift: Full Time | Day shift

Monthly Salary: ₹60,000 - ₹1,30,000 per month (Fixed + incentives)

Candidate Requirements: Graduate | Experienced Only | Good English

Joining Fee: No

Assessment Deadline: September 30, 2026

Role Overview

We are seeking a hybrid between a seasoned software architect and an AI whisperer—an engineer who understands that the future of software development relies on autonomous workflows and "vibe coding." You will not just be writing boilerplate code line-by-line; you will be orchestrating multi-agent systems, integrating the Model Context Protocol (MCP), and directing AI agents to write, test, and ship software alongside you.

Your immediate mandate is to move beyond basic chatbot APIs and build self-iterating, app-aware AI systems that can independently plan sequences of actions, call deterministic business-logic tools, evaluate their own outputs, and achieve complex objectives. Specifically, you will be instrumental in developing the Loopnow CPA Copilot, an AI-powered Canadian bookkeeping and GST/HST compliance agent.

Core Responsibilities

Agentic System Architecture: Design, test, and deploy stateful, multi-agent workflows (using LangGraph, CrewAI, AutoGen, or CopilotKit) to automate complex, multi-step business logic, specifically focusing on app-aware, bidirectional state synchronization between UI and agent.

Frontend & Full-Stack Integration: Lead frontend development and backend integration to ensure seamless data flow between the AI tools and the user interface.

Deterministic Rules Engines: Build robust, testable compliance layers where the LLM is not the source of truth. You will develop explicit rules engines for Canadian tax compliance (e.g., ITC calculations, CRA documentation tiers, GIFI mappings, and meals/entertainment rules).

Strict Tool Contracts: Define and implement deterministic tools using strict schemas (e.g., Zod). Ensure the AI relies on these backend tools for all calculations, classifications, and state transitions, returning structured JSON results.

AI-Native "Vibe Coding": Leverage AI coding agents and CLI harnesses (Cursor, Aider, Claude Code, OpenCode) to rapidly prototype, build, and ship production-ready applications at 10x traditional speed.

Streaming & Observability: Implement real-time streaming of token generation, tool calls, and state updates to ensure the UI visually exposes agent activity. Create robust audit trails for every meaningful agent action.

Security & Prompt Injection Resistance: Design systems that treat user input (e.g., receipt descriptions, vendor notes) as untrusted data. Implement strict authorization boundaries, human-in-the-loop approval mechanisms, and safe error handling.

Tech Stack & Requirements

AI & Orchestration Mastery: Deep expertise in LLM API integration, function calling, structured outputs, and building multi-step agent loops. Proven experience in ensuring AI systems recognize ambiguity and appropriately flag items for human review.

The "Vibe Coder" Mindset: You prioritize architectural design and system-level thinking over syntax memorization. You are fluent in pair-programming with AI, structuring complex prompts, managing context boundaries, and critically debugging AI-generated logic errors.

Programming & Frameworks: Strong fundamentals in Python and Node.js/TypeScript. Experience with state management frameworks that support bidirectional agent/application state (like CopilotKit shared state/AG-UI). Familiarity with modern frontend ecosystems (Next.js, Tailwind CSS) for full-stack prototyping.

Backend & Environment: Hands-on experience with Linux server environments, containerization, and writing deterministic API endpoints specifically optimized for AI tool-calling.

Evaluation & Testing: Ability to build self-verification loops where the system deterministically validates AI proposals before confirming completion.

Interview Process & Compensation

We do not conduct traditional Data Structures & Algorithms (DSA) whiteboard interviews. We care exclusively about your architectural thinking and how rapidly you can build alongside AI.

Round 1: The AI Architecture Discussion (45 mins): A deep dive into your past agentic projects, model selection logic, prompt engineering frameworks, and approaches to separating deterministic logic from probabilistic LLM outputs.

Round 2: The "Vibe Coding" Pair Session (60 mins): A live, screen-share session where you use your preferred AI harness to scaffold, debug, and deploy a tool-calling script.

Round 3: The Technical Assessment (Deadline: September 30, 2026): You are required to complete the Production-Ready AI CPA & Bookkeeper (Canada) Assessment. This will evaluate your ability to build an app-aware, stateful agent that uses deterministic tools for GST/HST rules, GIFI mapping, and compliance verification. Submissions are due by September 30, 2026.

Round 4: Culture & Vision Fit (30 mins): Discussing rapid iteration, extreme ownership, and alignment with our high-velocity engineering culture.

Perks & Benefits: Flexible Working Hours and an aggressive learning budget for API credits and premium AI tooling.