Overview

LLM Applications & Agentic Systems · Full-time · Mid-level (2–4 years) · Remote (India) or on-site

Responsibilities

• Design and ship agentic applications end to end — planning, tool use, skill chaining, and multi-agent workflows — using LangChain, LangGraph, and related orchestration frameworks.
• Build and chain agent skills — develop reusable skills (tools, functions, and capabilities) and compose them into multi-step skill chains that solve complex analytical workflows reliably.
• Build LLM orchestration pipelines that integrate models with APIs, databases, and internal services, with sensible state management, retries, and fallbacks.
• Own the data behind the AI — source, clean, structure, and chunk data for retrieval-augmented generation (RAG); choose and tune embeddings and vector stores.
• Measure what you ship — design evals and test sets, track quality metrics, and apply basic statistics to separate real improvements from noise.
• Harden systems for production — observability, tracing, guardrails, latency and cost optimization (model routing, caching, prompt design).
• Collaborate and communicate — work with product, data, and engineering peers to turn ambiguous problems into shippable AI features.

Qualifications

• 2–4 years of software or ML engineering experience, including hands-on experience building LLM-powered applications.
• Strong Python and experience with LLM orchestration frameworks — LangChain and LangGraph specifically, or deep experience with comparable tools (LlamaIndex, Semantic Kernel, CrewAI, AutoGen).
• Strong data understanding — comfortable with SQL, data pipelines, data quality, embeddings, and preparing messy real-world data for retrieval and model consumption.
• Working understanding of machine learning fundamentals: how models are trained and evaluated, common failure modes, and when to use (and not use) an LLM.
• Basic statistics: distributions, sampling, significance testing, and enough rigor to design and interpret A/B tests and offline evals.
• Cloud deployment experience (AWS/GCP/Azure), containerization, and CI/CD for ML systems.

Nice to have

• Production RAG experience with vector databases (pgvector, Pinecone, Qdrant, Weaviate) and hybrid search.
• Familiarity with LLM observability and eval tooling (LangSmith, Braintrust, Helicone) and prompt-caching mechanics.