Opportunity brief
About the Company: Operonn is an early-stage, founder-led AI company dedicated to helping businesses improve their ROI using AI. We focus on delivering AI solutions to industries and businesses to streamline operations and enhance revenue, moving beyond simple chatbots to create impactful production AI systems. Responsibilities: LLM application development: Build and maintain components of our agent and orchestration layer — tool routing, structured outputs, schema-validated generation, prompt versioning. Retrieval engineering: Implement and tune RAG pipelines over operational data — chunking, embedding, reranking, provenance tracking. Measure retrieval quality with honest metrics. Evaluation infrastructure: Build golden sets, regression harnesses, and LLM-as-judge frameworks. Catch regressions before they reach customers. Observability & telemetry: Instrument latency, cost, and quality metrics per LLM call. Trace failures end-to-end. Integration support: Contribute to ERP-side connectors and ingestion pipelines — FastAPI services, async workers, event handling. Documentation: Maintain technical documentation for every component you ship — readable by engineers who join after you. Code review participation: Review peer PRs, defend your own, take feedback that makes the system better. Requirements: Must-Have: Strong Python with hands-on experience in async patterns and production-grade code FastAPI or an equivalent Python web framework At least one shipped end-to-end LLM project — RAG, agent, fine-tune, or applied ML system that survived real users Working knowledge of vector databases (Qdrant, pgvector, or Weaviate) and embedding models Familiarity with at least one orchestration framework: LangChain, LangGraph, or LlamaIndex Git fluency, Docker awareness, Linux comfort Demonstrable ability to read technical papers and apply only what matters Good-to-Have: Open-source contributions of any scale TypeScript or Next.js for cross-stack work Cloud exposure — AWS (Lambda, ECS, RDS), GCP (Cloud Run, BigQuery), or Firebase Distributed systems or event-driven architecture background Fine-tuning, PEFT, or domain-adaptation experience Exposure to enterprise software environments — ERP, CRM, or ticketing systems Eligibility: Final-year BTech, recent graduate, or Masters student in Computer Science, AI, or a related field Self-taught engineers with verifiable portfolios are equally welcome Available 20–25 hours per week for a minimum 3-month commitment