Company Description BuildsenseAI is a Prop-Tech company revolutionizing the real estate industry through innovative artificial intelligence and intelligent automation solutions. The company focuses on transforming manual operations, fragmented workflows, and reactive processes into smart, data-driven decision-making systems. By integrating real estate, AI, and automation, BuildsenseAI develops platforms and tools to streamline operations in sales, customer engagement, compliance, workflows, and business intelligence for modern real estate businesses. BuildsenseAI is committed to empowering the real estate sector with cutting-edge technology for a more efficient future. Role Description This is a full-time, on-site role for an Artificial Intelligence Engineer based in Gurugram. The Artificial Intelligence Engineer will be responsible for designing, building, and optimizing AI-driven solutions for real estate applications, including implementing machine learning models, developing algorithms for pattern recognition and natural language processing, and collaborating with cross-functional teams to integrate AI functionality into core software products. The role involves extensive software development and leveraging cutting-edge AI technologies to drive innovation and enhance operational processes. What You Will Do Technical Architecture — You Own This Completely •Design the multi-agent orchestration layer: choose and implement the right framework (LangChain, LangGraph, AutoGen, or CrewAI) for PropOS's specific agent coordination requirements. This is the most critical architectural decision of the company's first year. •Build the WhatsApp-first delivery layer: integrate Meta WhatsApp Business API via a BSP partner (Gupshup or Kaleyra), handle multi-turn conversation state, manage session windows, implement media handling, and build robust fallback flows for API failures. •Design the shared data ontology: model the PropOS unified data layer: Land Parcel → Approvals → Units → Bookings → Collections → Handover. This schema determines whether cross-agent intelligence is possible — get it wrong and everything is siloed. •Own the LLM strategy: evaluate foundation models (GPT-4o, Claude, Sarvam AI, Krutrim) against cost, latency, and Hindi language quality benchmarks. Make the final call on which models power which agents. •Build document ingestion pipelines: parse RERA approval letters, BOQ Excel files, CP agreements, and project drawings using pypdf, camelot, Textract, and custom parsers. AI Agent Engineering — The Core Daily Work •Implement production LLM agents: with tool use, memory management, multi-step reasoning, and human-in-the-loop decision points. Agents must handle ambiguous real estate conversations, not just structured commands. •Build vernacular NLP capabilities: implement Hindi, Hinglish, and eventually Marathi/Gujarati intent classification. Evaluate Sarvam AI and IndicBERT for vernacular understanding. Fine-tune on real estate domain vocabulary. •Design and maintain the agent evaluation framework: define success metrics for each agent (task completion rate, hallucination rate, fallback frequency), build automated test suites, and run weekly quality reviews. •Set up LLM observability: implement LangSmith or equivalent for prompt tracking, latency monitoring, cost per conversation, and failure analysis. You will look at these dashboards every morning. Implement RAG pipelines: for developer project document search using Pinecone or Weaviate. Semantic chunking, re-ranking, and retrieval quality evaluation are your responsibility.