Senior AI Engineer

Job Title: Senior AI Engineer (Agentic AI, RAG & Knowledge Graphs)

Location: Hybrid / On-site (Bangalore, India)

Company: Spheresmith

Experience Level: Senior (5+ years in AI/ML & Software Engineering)

Spheresmith is a Decision Intelligence company that builds governed, accurate, and auditable decision systems aligned with each client's policies and guidelines. Originating in investment management for LPs, GPs, and family offices, Spheresmith extends its platform across corporate finance, sales, marketing, and IT.

By combining domain ontologies, reusable decision records, and private, domain-fine-tuned LLMs, our product suite—including Investsphere, Prism, and Corpsphere—ensures sensitive data stays within customer environments while converting ad hoc choices into transparent, traceable decision lifecycles. Over 35 enterprise customers rely on Spheresmith to give every high-impact decision clear context, ownership, and supporting data.

We are seeking a Senior AI Engineer to drive the architecture, development, and deployment of our next-generation decision intelligence systems. In this role, you will build autonomous and semi-autonomous multi-agent workflows, integrate structured domain ontologies with unstructured data via GraphRAG, fine-tune private domain-specific LLMs, and implement robust tool integration protocols like Model Context Protocol (MCP).

You will work at the intersection of production software engineering, cutting-edge agent frameworks, and enterprise-grade data science to deliver transparent, reproducible AI outputs for high-stakes business decisions.

• Agentic Architecture & Tool Orchestration: Design, implement, and optimize stateful multi-agent systems using LangGraph and LangChain. Build tool integration mechanisms via Model Context Protocol (MCP) and function calling to connect agents seamlessly with private enterprise databases, APIs, and microservices.
• Knowledge Graphs & Ontologies: Partner with domain experts to translate corporate guidelines and investment policies into formal domain ontologies. Integrate Knowledge Graphs (KGs) with vector retrieval (GraphRAG) to enable precise, auditable, and context-aware reasoning.
• Domain Fine-Tuning & Evaluation: Fine-tune open-weight LLMs (e.g., Llama 3, Mistral, Qwen) using parameter-efficient techniques (PEFT, LoRA/QLoRA) to run within private cloud environments. Establish rigorous evaluation frameworks (Ragas, TruLens, DeepEval) to guarantee accuracy and compliance.
• Advanced RAG Pipelines: Architect hybrid retrieval systems (dense vectors + sparse search + knowledge graphs) with re-ranking, query transformation, and strict citation tracking to eliminate hallucination.
• Data Science & Analytics: Conduct data analysis, embedding analysis, and model performance profiling to optimize retrieval accuracy, inference latencies, and output consistency across enterprise use cases.
• Enterprise AI Integration: Write production-grade, asynchronous Python code with Pydantic schemas, FastAPI endpoints, and clean microservice architecture. Deploy scalable workloads to cloud infrastructure (GCP/AWS) utilizing Docker/Kubernetes containerization.

• Experience: 5+ years of experience in AI/ML software engineering with a strong track record of deploying LLM-powered applications into production environments.
• Agentic Frameworks: Hands-on mastery of LangGraph, LangChain, LlamaIndex, or custom multi-agent orchestration engines.
• Model Context Protocol & APIs: Deep understanding of MCP, standard function-calling paradigms, REST/gRPC APIs, and tool execution boundaries.
• RAG & Knowledge Graphs: Proven experience building advanced RAG systems. Hands-on experience with graph databases (Neo4j, Memgraph, RDF/OWL ontologies, networkx) and GraphRAG methodologies.
• LLM Fine-Tuning: Practical experience fine-tuning LLMs using Hugging Face tools, Unsloth, LoRA/QLoRA, and dataset curation/synthetic data generation.
• Core Engineering & Data Science: Expert-level Python (asyncio, Pydantic, FastAPI) alongside solid data science foundations (embeddings, vector indexing with FAISS/Qdrant/Pinecone, statistics, data preprocessing).
• Databases & Vector Stores: Hands-on work with vector engines (Weaviate, Qdrant, PGVector) and relational/document databases.