Software Engineer

Company: Arionix

Website: Visit Website

Business Type: Startup

Company Type: Product

Business Model: B2B

Funding Stage: Pre-seed

Industry: Information Technology

Salary Range: ₹ 15-45 Lacs PA

Job Description

What You'll Do

• Design and evolve the ontology/schema for the enterprise context graph — entities (people, teams, systems, products, projects, customers, policies), relationships, and metadata that reflect how the organization operates.
• Build ingestion pipelines that extract structured context from source systems (CRM, ticketing, HRIS, wikis, code repos, data warehouses) and unstructured context (documents, Slack/Teams, meeting transcripts) using a mix of ETL, NLP, and LLM-based extraction.
• Implement entity resolution and deduplication so the same person, system, or concept is correctly linked across disparate sources.
• Architect the storage and query layer (graph database, hybrid graph + vector store) for low-latency retrieval by both applications and AI agents.
• Build APIs and retrieval interfaces (including MCP servers or similar protocols) so LLM agents and internal tools can query the graph for grounded, up-to-date context.
• Establish freshness, versioning, and provenance mechanisms — every fact in the graph should be traceable to a source and timestamp, and stale or conflicting data should be surfaced, not silently trusted.
• Partner with security and data governance teams to enforce access control at the entity/relationship level, since context graphs often surface sensitive cross-departmental information.
• Define and track quality metrics for the graph — coverage, accuracy, staleness, resolution rate — and instrument pipelines to catch regressions.
• Collaborate closely with applied AI/ML teams building agents and copilots that consume the graph, iterating based on what actually improves grounding and reduces hallucination.
  
  

• 5+ years in data engineering, knowledge engineering, or backend systems, with real experience shipping production data pipelines at scale.
• Hands-on experience with graph databases (Neo4j, Amazon Neptune, TigerGraph, or similar) and graph query languages (Cypher, Gremlin, SPARQL).
• Solid understanding of ontology and taxonomy design — RDF/OWL familiarity is a plus, but practical schema judgment matters more than academic purity.
• Experience building entity resolution / record linkage systems (rule-based, ML-based, or LLM-assisted).
• Strong Python (or equivalent) skills, with experience orchestrating pipelines (Airflow, Dagster, or similar) and working with both batch and streaming data.
• Practical experience integrating LLMs for extraction and structuring — turning unstructured text into structured graph facts, including handling ambiguity and low-confidence extractions.
• Familiarity with vector databases and hybrid retrieval (combining graph traversal with semantic search) for RAG-style applications.
• Understanding of data governance, access control, and provenance tracking in a multi-source enterprise environment.
• Comfort operating in ambiguity — this role often means defining the schema and the pipeline architecture from scratch, not implementing a pre-built spec.
  
  

• Experience with Model Context Protocol (MCP) or building tool/context interfaces for LLM agents.
• Background in semantic web technologies, knowledge representation, or search relevance engineering.
• Prior experience at a company that built an internal "single source of truth" system (data catalog, CMDB, master data management).
• Experience with change-data-capture (CDC) pipelines for keeping a graph continuously fresh.