SDE Internship | IIT/NIT students | AI-agents development

We are building full stack AI-agentic teams (hyper-specialised autonomous agents that use LLM back- ends, MCP servers, tailor-made tools and retrieval pipelines) to completely replace desk-job based businesses which currently require high-value ($100k+) human employees. We invite IIT/NIT students (ideal for 2025/26 batches) to join us as Software-Engineering Interns and help design, build, and benchmark these agents at production scale. 1 | Internship Snapshot • Role : Software Engineer Intern – LLM / Autonomous-Agent Development • Duration : 10–12 weeks (May–July 2025, flexible), can extend basis mutual fit • Location : Remote (IST overlap) • Stipend : Competitive + PPO-track (get to be a part of the founding team with ESOPs) • Stack: Python, FastAPI, LangChain/LangGraph, pgvector/FAISS, Docker, MCP servers • Work directly with founders (IIT/IIM alumni) 2 | What You Will Build • LLM-backed micro-agents that own an entire sub-task/workflow end-to-end • Retrieval pipelines using hybrid vector + keyword search (FAISS / pgvector) for tool- augmented reasoning. • Agent evaluation harnesses (pytest + LangChain), echoing best practice from industrial agent teams 3 | Core Responsibilities • Spec & API design – Draft OpenAPI contracts and JSONSchema for new agents. • Build agent functionalities: Integrate sessions, memory, knowledge graphs, state etc. • Build tools – Write typed, test-covered Python; integrate async I/O and streaming where it improves token efficiency • Infrastructure – Containerise services, and deploy to our EKS staging cluster with auto- scaling triggers. • Documentation & handover – Produce clear READMEs 4 | Ideal Candidate Must-Have • Solid CS fundamentals (DSA, OS, networks) • Fluency in Python; Git workflows • Experience with agent development on Agno or OpenAI SDKs • SQL + basic schema design Nice-to-have • REST, MCP and Knowledge graphs • AWS basics (S3, EKS) 5 | How to apply • Please fill this form: https://forms.gle/gjNN9bBSVMTmqs6aA Show more Show less