Role Overview :

Youll work alongside senior engineers and the founding team, building the AI systems behind the platform LLM pipelines, retrieval, and the production services around them.

What youll do :

- Build and maintain Python backend services and RESTful APIs using FastAPI; own your features through to production.

- Ship LLM-powered features RAG, prompt engineering, structured outputs, and tool calling using LLM APIs (OpenAI, Anthropic, Gemini).

- Build agentic and multi-step workflows with LangChain and LangGraph chains, tools, memory, and state.

- Build and maintain the retrieval layer : document ingestion, chunking, embedding, and querying a vector store (pgvector, Pinecone, or equivalent).

- Write and tune the SQL behind your features (MySQL, PostgreSQL), and add caching where it earns its place.

- Track how your AI features behave in production build evaluation sets, iterate on prompts, and watch latency and cost.

- Deploy and monitor your services on AWS (EC2, S3, RDS, Lambda), and use message queues for async work.

- Learn fast through code reviews with senior engineers, and document what you build.

What were looking for :

- Bachelors degree in CS, Engineering, or a related field.

- 1-2 years of hands-on Python development with working knowledge of FastAPI; strong internship experience counts.

- Strong Python fundamentals data structures, async/await, type hints, and clean modular code.

- Youve built at least one RAG pipeline end to end and can walk us through what broke and how you fixed it.

- Hands-on with LangChain and LangGraph chains, tools, memory, and stateful multi-step agentic workflows.

- Practical grip on LLM APIs (OpenAI, Anthropic, Gemini) prompt engineering, structured outputs, function and tool calling, streaming, and token/cost awareness.

- Hands-on with a vector store (pgvector, Pinecone, or equivalent) chunking strategies, embedding models, and similarity search.

- Working knowledge of SQL databases (MySQL or PostgreSQL) schema design, joins, and writing efficient queries. Redis for caching is a plus.

- Comfortable evaluating and debugging AI output building test sets, tracing failures, and telling whether the problem is the prompt, the retrieval, or the model.

- Comfortable using AI-assisted development tools (Cursor, Claude Code) to improve development speed and code quality.

Bonus points :

- Self-hosted models (vLLM, Ollama), fine-tuning (LoRA, QLoRA), MLOps tooling, message queues, or open-source and hackathon work with LLMs.

- Strong problem-solving skills and comfortable working independently in a fast-moving startup. Send links to projects or repos if you have them.