AI Engineer (Custom Agent Architecture)

Job description

Position: AI Engineer

Company Name: Ztrios Technology Ltd.

Address: House:2, Road:2/A, Sector:12, Uttara, Dhaka-1230

Job Type: Full-Time (On-Site)

About the Role

We build custom AI agents and embed them into our clients' systems. We're looking for an AI Engineer who goes beyond off-the-shelf frameworks: someone who can design agent architectures from first principles, adapt or rebuild them for each client's needs, and understands deeply how LLM-based systems behave in production.

Responsibilities

- Design and build custom AI agents end-to-end, from architecture to deployment

- Create, modify, and re-architect agent systems (planning loops, memory layers, tool routing, multi-agent orchestration) to fit specific business requirements

- Decide when to use a framework and when to build custom, and justify those trade-offs

- Integrate agents into existing systems (CRMs, ERPs, internal tools, legacy APIs)

- Build and maintain tool/function-calling layers, MCP servers, and API connectors

- Design context management, retrieval (RAG), and long-term memory strategies

- Set up evaluations, guardrails, and observability to keep agents reliable and safe

- Optimize agents for latency, cost, and accuracy across different models

- Stay current with research and new model capabilities, and bring them into our products

- Guide the team on agent design patterns and review architectural decisions

Requirements

- Strong software engineering skills in Python and/or TypeScript

- Proven track record of building production LLM agents (Claude, OpenAI, Gemini, or open-source models)

- Deep understanding of agent architectures: ReAct, planner–executor, hierarchical and multi-agent systems, state machines/graphs

- Ability to build agents without relying on a framework, plus hands-on experience with LangGraph, CrewAI, AutoGen, or similar

- Strong knowledge of prompt and context engineering, tool calling, and structured outputs

- Experience with RAG pipelines, embeddings, and vector databases

- Solid grasp of LLM limitations: hallucination, context limits, prompt injection, failure recovery

- Strong skills in REST APIs, webhooks, databases (SQL/NoSQL), Git, and cloud deployment

- Able to own a solution independently, from design through delivery

Nice to Have

- Experience building MCP servers or clients

- Fine-tuning, model distillation, or running open-source models (vLLM, Ollama, etc.)

- Agent evaluation frameworks and LLM observability tools (LangSmith, Langfuse, etc.)

- AI security knowledge (prompt injection defense, sandboxing, permission design)

- Docker, Kubernetes, CI/CD, and AWS/GCP/Azure

- Published work, open-source contributions, or a GitHub showing custom agent systems

What We Offer

- Architect real agentic AI products used by real clients

- High ownership in a small, fast-moving team

- Competitive salary based on experience

- A clear path to Lead AI Engineer / AI Architect

- 2 Festival Bonuses per year

- Annual Performance Bonus (Based on Performance)

- Salary review every 4 months(Based on Performance)

- Friendly, fast-growing work culture

- Growth-oriented environment with continuous learning opportunities

Salary

Negotiable

How to Apply

Send your CV to hr@ztrios.com with the subject line “AI Engineer for Ztrios – Your_Name”

Responsibilities

• Design and build custom AI agents end-to-end, from architecture to deployment
• Create, modify, and re-architect agent systems (planning loops, memory layers, tool routing, multi-agent orchestration) to fit specific business requirements
• Decide when to use a framework and when to build custom, and justify those trade-offs
• Integrate agents into existing systems (CRMs, ERPs, internal tools, legacy APIs)
• Build and maintain tool/function-calling layers, MCP servers, and API connectors
• Design context management, retrieval (RAG), and long-term memory strategies
• Set up evaluations, guardrails, and observability to keep agents reliable and safe
• Optimize agents for latency, cost, and accuracy across different models
• Stay current with research and new model capabilities, and bring them into our products
• Guide the team on agent design patterns and review architectural decisions

Requirements

• Strong software engineering skills in Python and/or TypeScript
• Proven track record of building production LLM agents (Claude, OpenAI, Gemini, or open-source models)
• Deep understanding of agent architectures: ReAct, planner–executor, hierarchical and multi-agent systems, state machines/graphs
• Ability to build agents without relying on a framework, plus hands-on experience with LangGraph, CrewAI, AutoGen, or similar
• Strong knowledge of prompt and context engineering, tool calling, and structured outputs
• Experience with RAG pipelines, embeddings, and vector databases
• Solid grasp of LLM limitations: hallucination, context limits, prompt injection, failure recovery
• Strong skills in REST APIs, webhooks, databases (SQL/NoSQL), Git, and cloud deployment
• Able to own a solution independently, from design through delivery
• Experience building MCP servers or clients
• Fine-tuning, model distillation, or running open-source models (vLLM, Ollama, etc.)
• Agent evaluation frameworks and LLM observability tools (LangSmith, Langfuse, etc.)
• AI security knowledge (prompt injection defense, sandboxing, permission design)
• Docker, Kubernetes, CI/CD, and AWS/GCP/Azure
• Published work, open-source contributions, or a GitHub showing custom agent systems

Benefits

• Architect real agentic AI products used by real clients
• High ownership in a small, fast-moving team
• Competitive salary based on experience
• A clear path to Lead AI Engineer / AI Architect
• 2 Festival Bonuses per year
• Annual Performance Bonus (Based on Performance)
• Salary review every 4 months(Based on Performance)
• Friendly, fast-growing work culture
• Growth-oriented environment with continuous learning opportunities

Skills

• LangGraph
• LLM Agents
• Python
• RAG
• REST APIs
• TypeScript
• Vector Databases