Website:
butechnologies.co
Job details:
About the Role
We are looking for an Agentic AI Engineer to design, build, and operate autonomous AI systems that can plan, reason, and take multi-step actions to accomplish complex goals. In this role you will move beyond traditional ML pipelines to architect AI agents that interface with tools, APIs, code environments, and real-world data — with a strong focus on reliability, safety, and human-in-the-loop oversight.
You will work at the frontier of applied AI, shipping systems that leverage the latest foundation models while building the scaffolding, evaluation frameworks, and infrastructure needed to make them production-ready.
Key Responsibilities
• Design and implement agentic AI systems using frameworks such as LangGraph, AutoGen, CrewAI, or custom orchestration layers built on top of foundation model APIs.
• Architect multi-agent pipelines that decompose complex tasks into sub-goals, delegate to specialist agents, and synthesize results reliably.
• Build and maintain tool-use layers — including web search, code execution sandboxes, database connectors, and REST/GraphQL API integrations — that agents can invoke safely.
• Develop robust prompt engineering strategies, chain-of-thought scaffolding, and retrieval-augmented generation (RAG) pipelines to maximise model reasoning quality.
• Implement comprehensive evaluation and observability frameworks (evals, traces, human feedback loops) to measure agent performance, failure modes, and regression.
• Define and enforce safety guardrails, output validation, and human escalation triggers to ensure responsible autonomous behaviour in production.
• Collaborate with product, design, and domain experts to translate business requirements into agentic workflows and ship iteratively.
• Stay current with rapidly evolving research (ReAct, Reflexion, Constitutional AI, function-calling advances) and assess applicability to product goals.
• Contribute to internal tooling, shared libraries, and best practices that accelerate the broader engineering organisation's use of AI agents.
Required Qualifications
• 5+ years of software engineering experience, with at least 2 year focused on LLM-based systems or AI agent development.
• Strong Python proficiency; comfort writing production-quality async code, REST clients, and data pipelines.
• Hands-on experience with one or more major LLM providers (OpenAI, Anthropic, Google Gemini, open-source models via HuggingFace / Ollama).
• Proven ability to build and debug multi-step agentic workflows including tool-calling, memory management, and context window optimisation.
• Solid understanding of RAG architectures, vector databases (Pinecone, Weaviate, pgvector), and embedding strategies.
• Experience designing evaluation harnesses and working with human-in-the-loop feedback systems.
• Familiarity with cloud platforms (AWS, GCP, or Azure) and containerised deployment (Docker, Kubernetes).
• Strong analytical mindset with a bias toward empirical testing over intuition.
Preferred Qualifications
• Experience with agent frameworks such as LangGraph, AutoGen, Semantic Kernel, CrewAI, or Haystack.
• Familiarity with the Model Context Protocol (MCP) or similar tool-use standards.
• Background in reinforcement learning from human feedback (RLHF) or preference optimisation.
• Contributions to open-source AI projects or published writing on agent architectures.
• Experience operating AI systems in regulated or safety-critical environments.
• Comfort working in a fast-paced, research-adjacent environment with high ambiguity.
Representative Tech Stack
Languages - Python (primary), TypeScript
LLM APIs - Anthropic Claude, OpenAI GPT-4o, Google Gemini
Agent Frameworks - LangGraph, AutoGen, CrewAI, custom orchestration
Vector / Memory - Pinecone, pgvector, Redis, in-context summarisation
Observability - LangSmith, Weights & Biases, OpenTelemetry, custom eval harnesses
Infrastructure - AWS / GCP, Docker, Kubernetes, GitHub Actions
Data - PostgreSQL, S3, dbt, Apache Kafka
Click on Apply to know more.