AI Agent Engineer

About Nourma We're building the AI-powered finance operating system that transforms how companies manage their financial operations. Our Decision Intelligence platform combines LLMs, multi-agent systems, and real-time data integration to create an intelligent finance team The Role We're seeking an AI/ML Engineer with deep expertise in LangChain, LlamaIndex, PydanticAI, and modern Python frameworks to architect and build the core intelligence layer of Nourma. Key Responsibilities LLM Orchestration & RAG Development (LangChain/LlamaIndex/PydanticAI Focus) Architect complex LangChain pipelines for multi-agent financial workflows Build production RAG systems using LlamaIndex for financial document retrieval Implement agents with strong type safety and structured outputs Design and implement: Chain-of-thought reasoning for financial analysis Dynamic prompt routing based on query complexity Memory management for long-running financial conversations Tool integration for agents to access GL, bank feeds, and operational data Optimise token usage and response latency for real-time WhatsApp interactions API Development & Integration (FastAPI Focus) Build high-performance FastAPI services for: Agent-to-agent communication protocols WhatsApp webhook processing with sub-second response Real-time financial data APIs for frontend consumption Design GraphQL schemas for flexible financial data queries Implement WebSocket connections for live financial updates Create robust error handling and retry mechanisms for financial integrations Vector Database & Semantic Search (Chroma Focus) Design and optimise Chroma collections for: Financial document embeddings (loan agreements, invoices) Conversation history and context retrieval Business logic and rule storage Implement hybrid search combining vector similarity and metadata filtering Build embedding pipelines for various document types (PDFs, emails, chat logs) Infrastructure & Scalability Deploy and manage LLM applications. Implement Redis caching strategies for LLM responses and financial data Design microservices architecture for agent deployment Set up monitoring and observability for AI pipelines Technical Requirements Must Have - Core Technologies Expert-level proficiency in: LangChain : Custom chains, agents, tools, memory systems LlamaIndex : Document stores, indices, query engines PydanticAI : Agent frameworks, type-safe LLM interactions, structured outputs FastAPI : Async programming, dependency injection, middleware Strong experience with Python async/await patterns Production experience with Chroma or similar vector databases Proficiency with Redis for caching and session management Experience with data pipeline and storage tools (Kafka, Spark, Airflow) for building scalable systems Nice to Have Knowledge of PostgreSQL and BigQuery for analytical workloadsUnderstanding of financial data structures (journal entries, chart of accounts) Experience with financial APIs (QuickBooks, Xero, Plaid, banking APIs) Knowledge of data consistency requirements for financial systems GraphQL schema design and optimisation Experience with WhatsApp Business API Background in fintech or accounting software Tech Stack LLMs : GPT-4, Claude, open-source models ML/AI : LangChain, LlamaIndex, PydanticAI, PyTorch, Transformers Vector DB : Chroma Data : PostgreSQL, BigQuery, Apache Kafka, Spark, Airflow APIs : FastAPI, GraphQL Infrastructure : AWS/GCP, Kubernetes, Docker, Redis Monitoring : Prometheus, Grafana, OpenTelemetry What We Offer Work on cutting-edge problems combining LLMs with real-time financial data Build systems processing millions of financial transactions Direct impact on how thousands of companies manage finances Work directly with founders and shape technical direction  Ideal Candidate Profile You're excited about: Building production LangChain and PydanticAI applications at scale Creating high-performance APIs that power AI agents Designing scalable architectures for financial data processing Working with cutting-edge LLM technologies You've probably: Built production LangChain/LlamaIndex/PydanticAI applications serving 1000+ users Created FastAPI services handling high-throughput LLM requests Worked with vector databases in production environments Designed data processing pipelines for financial or similar domains Contributed to open-source AI/ML projects Show more Show less