Full-Stack Developer — AI Automation SaaS Platform | Python + Chrome Extension + LLM Integration | 4-Week Sprint
Budget: ₹150,000 – ₹250,000 INR
Project Overview
I am building a B2B AI automation platform for small and medium businesses. The platform monitors email and calendar, surfaces intelligent summaries and recommended actions, automates repetitive GUI workflows across web applications, and gives teams an AI assistant accessible via a Chrome extension sidebar. This is a production build — not a prototype. The platform needs to be ready for real paying customers at the end of the engagement.
What You Will Build
The full product, end to end. This includes:
• A Python backend API deployed on a cloud platform, handling authentication, AI routing, cost management, and all business logic
• A Chrome Manifest V3 browser extension with a sidebar interface — the primary user surface
• Continuous email and calendar monitoring (Gmail, Outlook, Google Calendar) using IMAP IDLE and CalDAV polling
• An AI routing layer that selects the right LLM for each task based on complexity, routes through a cost guardrail stack, and falls back gracefully if a provider is unavailable
• A four-layer memory and knowledge recall system: local vector store, cloud vector store, long-term organisational memory, and a semantic knowledge graph
• A GUI automation system that records a user demonstrating a workflow once, then replays it across web applications using the browser accessibility tree — with an AI-powered self-healing mechanism when application UIs change
• A workflow scheduler that runs recorded automations automatically on triggers (time, email event, calendar event)
• An approval gate system — any automation step that sends, submits, or deletes anything requires explicit human approval before execution, with no exceptions
• Voice input across the extension and desktop using an on-device speech-to-text pipeline and a wake word
• Intelligent document handling — index uploaded files into the recall system and allow users to query them via the AI assistant
• 18 plugin integrations: Gmail, Outlook, Google Calendar, Outlook Calendar, WhatsApp Business, Google Drive, OneDrive, Google Docs/Sheets, Notion, Slack, Xero, QuickBooks, Salesforce (read), HubSpot (read), and custom webhook
• An admin console for the product owner: cost monitoring, kill switches, pricing overrides, and promo codes
• A demo organisation system for running prospect demonstrations
• A partner sub-console for reseller partners to manage their clients and run demos
Technology Stack
The stack is already decided. You are implementing within it, not choosing it. Key technologies:
• Backend: Python, FastAPI
• Database: PostgreSQL with vector extension, Redis for caching and counters
• Frontend / CDN: Cloudflare Pages, R2 storage, Cloudflare Workers
• Browser extension: Chrome Manifest V3
• LLM providers: Anthropic Claude, Qwen (Alibaba), Google Gemini — routed via a single adapter layer
• Memory: ChromaDB (local), pgvector, mem0, Graphiti knowledge graph with Neo4j
• Workflow orchestration: LangGraph
• GUI automation: Playwright, browser accessibility tree (primary), Claude computer_use API (visual fallback)
• Voice: Faster-Whisper (on-device STT), Pipecat, OpenWakeWord
• Payments: Stripe
• OAuth management: Nango
• Deployment: Railway
Timeline and Engagement Structure
• Duration: 4 weeks (28 calendar days from kick-off)
• This is an AI-assisted build. Claude handles architecture guidance; Codex handles code generation. Your role is direction, integration, testing, and deployment — not line-by-line authoring.
• Approximate total 120 hours of work for a well qualified developer with 5 plus years of experience having handled multiple projects in the past.
• There are four weekly milestone gates with clear exit criteria. Continuation to the next week is contingent on the previous gate being passed.
What I Am Looking For
• Strong Python backend experience — FastAPI, async patterns, database migrations
• Demonstrated experience with LLM APIs and AI application development (not just calling GPT — actual agentic pipelines)
• Chrome extension development experience (Manifest V3 specifically)
• Experience with Playwright or similar browser automation
• Able to work independently with a detailed specification — I provide architecture documents, you implement
• Clear communication in English — daily updates required
• Available for a 4-week focused sprint starting within 1 week of agreement
What I Am NOT Looking For
• Agencies that will subcontract this work to someone else without disclosure
• Developers who need daily hand-holding on standard implementation decisions
• Anyone who will push back on the technology stack — the stack is decided
• Bids that include a 'discovery phase' before any building starts — the specification is complete
How to Apply
Your bid must include the following or it will not be reviewed:
• A brief description of the most complex LLM / AI application you have built — what it did, what stack you used, what was hardest
• Your experience with Chrome Manifest V3 extensions specifically — not extensions in general
• Whether you have worked with LangGraph, mem0, or Graphiti before — and if not, your honest assessment of ramp-up time
• Your actual availability for the 4 weeks — start date and any blackout days
• One sentence on your approach to the single-LLM-adapter-file rule (all LLM calls through one file, no direct imports elsewhere) — this tells me immediately whether you have read the brief
Shortlisted candidates will be provided with a detailed project brief document before any call. A short technical assessment may be required before engagement. No NDA is required before the brief is shared — the brief is designed to give you what you need to assess the project without revealing proprietary details.
I am building a B2B AI automation platform for small and medium businesses. The platform monitors email and calendar, surfaces intelligent summaries and recommended actions, automates repetitive GUI workflows across web applications, and gives teams an AI assistant accessible via a Chrome extension sidebar. This is a production build — not a prototype. The platform needs to be ready for real paying customers at the end of the engagement.
What You Will Build
The full product, end to end. This includes:
• A Python backend API deployed on a cloud platform, handling authentication, AI routing, cost management, and all business logic
• A Chrome Manifest V3 browser extension with a sidebar interface — the primary user surface
• Continuous email and calendar monitoring (Gmail, Outlook, Google Calendar) using IMAP IDLE and CalDAV polling
• An AI routing layer that selects the right LLM for each task based on complexity, routes through a cost guardrail stack, and falls back gracefully if a provider is unavailable
• A four-layer memory and knowledge recall system: local vector store, cloud vector store, long-term organisational memory, and a semantic knowledge graph
• A GUI automation system that records a user demonstrating a workflow once, then replays it across web applications using the browser accessibility tree — with an AI-powered self-healing mechanism when application UIs change
• A workflow scheduler that runs recorded automations automatically on triggers (time, email event, calendar event)
• An approval gate system — any automation step that sends, submits, or deletes anything requires explicit human approval before execution, with no exceptions
• Voice input across the extension and desktop using an on-device speech-to-text pipeline and a wake word
• Intelligent document handling — index uploaded files into the recall system and allow users to query them via the AI assistant
• 18 plugin integrations: Gmail, Outlook, Google Calendar, Outlook Calendar, WhatsApp Business, Google Drive, OneDrive, Google Docs/Sheets, Notion, Slack, Xero, QuickBooks, Salesforce (read), HubSpot (read), and custom webhook
• An admin console for the product owner: cost monitoring, kill switches, pricing overrides, and promo codes
• A demo organisation system for running prospect demonstrations
• A partner sub-console for reseller partners to manage their clients and run demos
Technology Stack
The stack is already decided. You are implementing within it, not choosing it. Key technologies:
• Backend: Python, FastAPI
• Database: PostgreSQL with vector extension, Redis for caching and counters
• Frontend / CDN: Cloudflare Pages, R2 storage, Cloudflare Workers
• Browser extension: Chrome Manifest V3
• LLM providers: Anthropic Claude, Qwen (Alibaba), Google Gemini — routed via a single adapter layer
• Memory: ChromaDB (local), pgvector, mem0, Graphiti knowledge graph with Neo4j
• Workflow orchestration: LangGraph
• GUI automation: Playwright, browser accessibility tree (primary), Claude computer_use API (visual fallback)
• Voice: Faster-Whisper (on-device STT), Pipecat, OpenWakeWord
• Payments: Stripe
• OAuth management: Nango
• Deployment: Railway
Timeline and Engagement Structure
• Duration: 4 weeks (28 calendar days from kick-off)
• This is an AI-assisted build. Claude handles architecture guidance; Codex handles code generation. Your role is direction, integration, testing, and deployment — not line-by-line authoring.
• Approximate total 120 hours of work for a well qualified developer with 5 plus years of experience having handled multiple projects in the past.
• There are four weekly milestone gates with clear exit criteria. Continuation to the next week is contingent on the previous gate being passed.
What I Am Looking For
• Strong Python backend experience — FastAPI, async patterns, database migrations
• Demonstrated experience with LLM APIs and AI application development (not just calling GPT — actual agentic pipelines)
• Chrome extension development experience (Manifest V3 specifically)
• Experience with Playwright or similar browser automation
• Able to work independently with a detailed specification — I provide architecture documents, you implement
• Clear communication in English — daily updates required
• Available for a 4-week focused sprint starting within 1 week of agreement
What I Am NOT Looking For
• Agencies that will subcontract this work to someone else without disclosure
• Developers who need daily hand-holding on standard implementation decisions
• Anyone who will push back on the technology stack — the stack is decided
• Bids that include a 'discovery phase' before any building starts — the specification is complete
How to Apply
Your bid must include the following or it will not be reviewed:
• A brief description of the most complex LLM / AI application you have built — what it did, what stack you used, what was hardest
• Your experience with Chrome Manifest V3 extensions specifically — not extensions in general
• Whether you have worked with LangGraph, mem0, or Graphiti before — and if not, your honest assessment of ramp-up time
• Your actual availability for the 4 weeks — start date and any blackout days
• One sentence on your approach to the single-LLM-adapter-file rule (all LLM calls through one file, no direct imports elsewhere) — this tells me immediately whether you have read the brief
Shortlisted candidates will be provided with a detailed project brief document before any call. A short technical assessment may be required before engagement. No NDA is required before the brief is shared — the brief is designed to give you what you need to assess the project without revealing proprietary details.