Agentic AI Automation Engineer — Backend & Prompts

Nysonian; Islamabad Capital Territory; On-site; PKR 2,600,000 - 4,500,000; Full time

Job summary

Nysonian in Islamabad is seeking a Senior Machine Learning Engineer to build agentic AI systems that read data, decide, and take actions. You will design prompts, implement robust backend endpoints, and work with RAG, embeddings, and vector search to ground responses in real data.

You'll collaborate across product and engineering, iterate on prompts and models, and stay current with new model releases and tooling. The role is in-person in Islamabad with evening/night hours aligned to PKT.

Qualifications

• Solid hands on experience with agentic AI concepts: tool calling, multi step reasoning, memory, and orchestrating more than one agent to solve a problem.
• Practical experience with at least one agent framework such as LangChain, LangGraph, CrewAI, AutoGen, or the OpenAI or Anthropic agent/assistant SDKs.
• Real backend development experience in Python, with a framework such as FastAPI, Flask, or Django, including designing and deploying an API that other systems actually call
• A working understanding of NLP and how LLMs behave: context windows, tokens, embeddings, fine tuning versus retrieval, and why a model says confident things that are wrong
• Comfort with prompt engineering as an actual practice, not guesswork: iterating on a prompt, testing it against real cases, and knowing when the fix belongs in the prompt versus in the code around it
• A genuine foundation in data science and machine learning: you should be able to talk through a basic model training and evaluation workflow, know why accuracy is often the wrong metric, and be comfortable with Python’s data stack (pandas, NumPy, scikit-learn)
• At least one real project, professional or personal, that touches this space. We would rather see one thing you built end to end than a long list of tools you have briefly touched
• Genuine curiosity about the field. You should already be someone who reads about new model releases, checks benchmark leaderboards, and tries out new AI coding tools on your own, not because a job requires it

Responsibilities

• Building and maintaining agentic workflows that read customer data, make decisions, and take real actions (refunds, order updates, escalations, and similar)
• Writing, testing, and refining prompts, including thinking through where a model is likely to hallucinate or get overconfident, and designing around it
• Building and scaling backend endpoints (using FastAPI, Flask, or Django) that serve these AI systems reliably in production, not just in a notebook
• Working with retrieval systems (RAG), embeddings, and vector search to ground responses in real data instead of the model’s memory
• Reviewing and improving existing automations built in n8n, and helping decide when something should move from a no code workflow into real code
• Digging into a wrong or weird agent output, tracing it back through logs and executions, and fixing the actual root cause
• Staying current enough with the field that you naturally bring up a new model release, a new benchmark, or a new coding tool without being asked

Skills

Agentic AI concepts

Agent frameworks

Python backend

NLP fundamentals

Prompt engineering

Data science basics

End-to-end project

Curiosity about field

Tools

Python

FastAPI

Flask

Django

LangChain

LangGraph

CrewAI

OpenAI API

Anthropic API

n8n

Vector databases

Job description

Nysonian in Islamabad is seeking a Senior Machine Learning Engineer to build agentic AI systems that read data, decide, and take actions. You will design prompts, implement robust backend endpoints, and work with RAG, embeddings, and vector search to ground responses in real data.

You'll collaborate across product and engineering, iterate on prompts and models, and stay current with new model releases and tooling. The role is in-person in Islamabad with evening/night hours aligned to PKT.