Data Scientist - MID Profile - BN 12/11/26

MARGO
Warsaw, Poland
Hybrid

Who this role is best for

Strong fit for data scientists who architect agentic AI solutions and optimize LLM pipelines, with a focus on enterprise software and Investment Banking domain.

Best fit for

  • Candidates with experience in agentic AI and LLM pipeline optimization for enterprise finance applications
    — “Design and develop LLM-based agentic solutions to automate complex business processes.”
  • Individuals who have deployed open-weight models on GPU infrastructure with DevOps integration
    — “Deploy and industrialize models (LLMs, GPUs), ensuring quality through testing and DevOps best practices.”
  • Professionals who can balance technical innovation with compliance and auditability in regulated environments
    — “Dedication to shipping stable, secure code into flagship banking infrastructure where compliance, auditability, and data security are non-negotiable.”

Things to consider

  • Candidates must be prepared to work in a high-impact, strategic AI environment within Investment Banking
    — “A strategic environment within the platform, with high-impact use cases in document analysis, credit, and finance.”

How to stand out

  • Emphasize your experience with LangGraph or Google ADK in agentic graph design and implementation
    — “Architecting multi-step, stateful agentic graphs using LangGraph or Google ADK.”
  • Highlight your hands-on work with vLLM and GPU deployment for high-throughput LLMs
    — “Deploying and running high-throughput open-weight models on GPU infrastructure using runtimes such as vLLM.”
  • Showcase your expertise in testing ML/LLM pipelines with pytest and integration with DevOps tools
    — “Hands-on practice in pytest (unit, functional, and integration testing for ML/LLM pipelines).”
  • Demonstrate your ability to build robust safety rails and communicate LLM limitations effectively
    — “You remember to respect the boundaries and non-deterministic nature of LLMs, communicate constraints clearly, and build robust safety rails.”
  • Position yourself as someone who can quickly validate AI ideas with data and avoid technical debt
    — “You are testing ideas and PoCs quickly, validate feasibility with data and avoid technical debt.”
Pace · Fast PacedCollaboration · HighAutonomy · MediumDecision Impact · Team

Derived from job-description analysis by Serendipath's career intelligence engine.

What success looks like

  • Proven track record building asynchronous microservices and endpoints using FastAPI
  • Hands-on experience optimizing LLM pipelines programmatically using DSPy
  • Experience setting up tracing, drift detection, and automated evaluations for generative workflows using tools like Phoenix Arize
Typical background
Experience in enterprise software design patternsBackground in data science and machine learning

Skills & requirements

Required

PythonFastapiPytestLlms ArchitectureLanggraphDspyVllmKubernetesJenkinsGPU InfrastructureDevOps

Preferred

CUDAFine-tuningPhoenix Arize

Stack & domain

PythonFastapiPytestBashLlms ArchitectureLanggraphDspyVllmCUDAJenkinsPhoenix ArizeLeadershipCommunicationFinanceInvestment Banking

About the role

Original posting from MARGO via Lever

Context:

A team responsible for the development and industrialization of applications for the Investment Banking division. Accelerating the integration of generative AI solutions and AI agents into business processes. A strategic environment within the platform, with high-impact use cases in document analysis, credit, and finance.

Expectations:

Strong Python with extensive experience in Object-Oriented Programming and enterprise software design patterns.

Proven track record building asynchronous microservices and endpoints using FastAPI.

Hands-on practice in pytest (unit, functional, and integration testing for ML/LLM pipelines).

Bash scripts for automation, data handling, and container interaction.

LLMs Architecture understanding proven by project portfolio.

Architecting multi-step, stateful agentic graphs using LangGraph or Google ADK.

Hands-on experience optimizing LLM pipelines programmatically using DSPy.

Practical knowledge of deploying and running high-throughput open-weight models on GPU infrastructure using runtimes such as vLLM (experience with CUDA and fine-tuning is a strong plus).

Hands-on familiarity with automated build, test, and release pipelines (Jenkins).

Experience setting up tracing, drift detection, and automated evaluations for generative workflows using tools like Phoenix Arize.

Dedication to shipping stable, secure code into flagship banking infrastructure where compliance, auditability, and data security are non-negotiable.

General Stack: Google ADK, LangGraph, DSPy, vLLM, FastAPI, Python, Pytest. Arize Phoenix. Kubernetes, Jenkins, GPUs.

Responsibilities:

Design and develop LLM-based agentic solutions to automate complex business processes.

Work closely with business teams to identify needs and drive the adoption of AI solutions.

Develop data extraction pipelines and implement evaluation and observability frameworks for AI agents.

Deploy and industrialize models (LLMs, GPUs), ensuring quality through testing and DevOps best practices

You are taking full ownership in driving AI features.

You remember to respect the boundaries and non-deterministic nature of LLMs, communicate constraints clearly, and build robust safety rails.

You are a team player, communicate technical trade-offs cleanly with architects and business users, and enjoy mentoring peers.

You are testing ideas and PoCs quickly, validate feasibility with data and avoid technical debt.

Career Prospects:

Build expertise in Agentic AI architectures and the industrialization of LLMs in production. Work on high-visibility projects within the Investment Banking division.

Source: MARGO careers (Lever)

Similar roles

More roles at MARGO