New opportunity

Applied AI Scientist – AI Fellowship Program

Balyasny Asset Management · New York, United States

About this role

Balyasny Asset Management’s AI Fellowship is a rotational program placing applied AI talent with investment teams. The fellow begins with the central AI team, learning its in-house platform and delivering production tooling, then rotates onto investment teams as a dedicated AI partner to portfolio managers. The program aims to find a permanent home for the fellow on an investment desk.

Responsibilities

Design, build and deploy agentic workflows, data pipelines, research automation, screening tools and model tooling to improve portfolio managers’ research and decision speed.
Work with portfolio managers and analysts to turn investment workflows and ambiguous requirements into working technical solutions.
Contribute production tools to the central AI platform, work across multiple teams and strategies, and pilot emerging LLM techniques and agentic frameworks on live investment problems.

Required qualifications

3–5 years in a user-facing role scoping, building and deploying LLM-based solutions in production; experience building LLM applications with agentic frameworks and working directly with non-technical stakeholders.
A master’s degree in AI, Data Science, Computer Science, Statistics, Mathematics or a related quantitative field, or equivalent work experience. Strong grounding in computer science, statistics, machine learning and generative AI.
High proficiency in Python for engineering, statistical modelling and machine learning; experience integrating LLMs and AI models into production workflows; familiarity with model-provider APIs and agentic frameworks; fluency with agentic coding tools; and working knowledge of MCP for tool and data integration.
Strong analytical, problem-solving and communication skills; ability to explain technical trade-offs, work independently amid ambiguity, manage multiple workstreams and meet deadlines.

Preferred qualifications include an advanced degree, investment-setting exposure such as buy-side or equity research, and familiarity with cloud platforms, Git and workflow orchestration tools such as Airflow. For NY, CA and IL applicants, starting base pay is USD 150000–225000 annually, depending on factors including experience, business needs and market demands. Bonus compensation and employee benefits may also be available.

For full details about this role, click “Apply on company site” to view the employer’s original posting.

Skills for this role

PythonMachine LearningGenerative AILarge Language ModelsAgentic workflowsResearch automationData pipelinesStatistical modellingLangGraphOpenAI Agents SDKClaude Agent SDKClaude CodeCodexModel Context Protocol (MCP)OpenAI APIAnthropic APIGoogle AI APIsAWSAzureGCPGitAirflowStakeholder communicationProblem-solving

Checking your profile…

Preparation suggestions, not the employer’s actual interview questions. Always check the original posting for current requirements.

Understand the role

What matters most

Ship AI tools that speed investment research
The role designs and deploys agentic workflows, data pipelines, research automation, screening tools and model tooling for portfolio managers. Prepare a concise example of a production LLM solution, including what it changed in the user’s workflow.
Turn ambiguous desk needs into working solutions
The fellow works directly with portfolio managers and analysts to shape investment workflows into technical solutions. Prepare to explain how you scope a request with non-technical stakeholders and communicate implementation trade-offs.
Contribute to the central AI platform
The initial months are spent learning BAM’s in-house AI ecosystem while shipping production tooling to the central platform. Prepare to discuss how you integrate models, APIs and tools into an existing production workflow.
Adapt across teams and test new approaches
The fellowship rotates across investment teams and strategies and calls for piloting emerging LLM techniques on live problems. Prepare an example of evaluating a new approach before deciding whether to use it.

Likely challenges to think through

Balancing speed with dependable outputs
Preparation hypothesis, not a stated employer problem: tools intended to accelerate live investment research may need checks that slow delivery. Be ready to discuss how you would test output quality and handle failures before a portfolio manager relies on a workflow.
Balancing shared tooling with desk-specific needs
Preparation hypothesis: central-platform contributions and later desk rotations could pull solutions toward reuse or customization. Prepare to explain how you would decide what belongs in shared tooling versus a team-specific workflow.
Managing shifting priorities across rotations
Preparation hypothesis: multiple teams, ambiguous requests and deadlines may make prioritization difficult. Prepare an example of narrowing scope, agreeing on a useful first deliverable and managing concurrent workstreams.

Practice your answers

Technical practice

How would you design an agentic workflow to support a portfolio manager’s research?

What to coverThe role calls for agentic workflows that improve research and decision speed. Suggested preparation: describe task boundaries, tool use, human review points and how you would judge whether the workflow is useful.

How would you turn an ambiguous investment workflow into a technical design?

What to coverThe posting emphasizes translating portfolio managers’ and analysts’ workflows into working solutions. Suggested preparation: explain how you would map the current process, clarify the decision being supported and agree on a narrow first deliverable.

How would you build a reliable data pipeline for research automation?

What to coverData pipelines and research automation are stated responsibilities. Suggested preparation: discuss data validation, handling missing or changing inputs, reproducibility and how failures would become visible to users.

How would you assess whether a research screening tool is useful?

What to coverThe fellowship includes building screening tools for live investment problems. Suggested preparation: distinguish a technically correct screen from one that saves analyst time, and discuss checks for false positives, missed results and changing data.

Where would you use statistical modelling rather than an LLM?

What to coverThe role requires Python proficiency in statistical modelling and machine learning alongside generative AI. Suggested preparation: compare the approaches by task, available evidence, interpretability and validation needs; avoid assuming an LLM is always the better choice.

How would you evaluate an emerging LLM or agentic approach before piloting it?

What to coverThe posting asks fellows to track new techniques, pilot promising approaches and bring proven wins to teams. Suggested preparation: define a baseline, representative tasks, quality and speed measures, and a criterion for stopping or expanding the pilot.

What would you consider when integrating an LLM application into a production workflow?

What to coverProduction deployment and integration of LLMs and AI models are explicit requirements. Suggested preparation: cover testing, failure handling, monitoring, operating cost and when a user should review or override an output.

How would you compare model-provider APIs for a specific application?

What to coverThe posting names familiarity with OpenAI, Anthropic and Google APIs. Suggested preparation: explain how you would test providers against the application’s quality, latency, cost and integration needs rather than assuming one is universally best.

How would you use MCP to connect an agent to tools and data?

What to coverWorking knowledge of MCP for tool and data integration is listed. Suggested preparation: explain the integration boundary, what access the agent needs and how you would test tool behavior and handle errors.

How would you move a desk-specific tool toward reuse on the central AI platform?

What to coverThe fellow initially contributes production tooling to the central platform and later works across investment teams. Suggested preparation: discuss which components to generalize, what should remain specific to a team, and how you would avoid disrupting an existing workflow.

Behavioural practice

Tell me about a time you scoped a solution with a non-technical stakeholder.

What to coverThe role involves working directly with portfolio managers and analysts. Suggested preparation: use a real situation, the questions and decisions you drove, and the resulting deliverable or feedback.

Describe a time you explained a difficult technical trade-off.

What to coverThe posting explicitly calls for explaining trade-offs to portfolio managers and analysts. Suggested preparation: give the situation, how you made the options understandable, what was decided and what happened.

Tell me about a time you delivered amid ambiguity and limited supervision.

What to coverSelf-direction and a bias toward shipping are stated expectations. Suggested preparation: describe the uncertain starting point, how you set scope and checked assumptions, and the result.

Describe how you managed competing workstreams against deadlines.

What to coverThe role calls for managing multiple workstreams while delivering on time. Suggested preparation: give a real example of prioritization, communication about constraints, actions taken and the outcome.

Tell me about adapting your approach for a different team or problem.

What to coverThe fellowship rotates across investment teams and strategies. Suggested preparation: choose a real example, explain how you learned the new context and changed your approach, and state the result without implying prior investment experience is required.

Sharpen your knowledge

Topics to prepare

Production LLM integration
Posting: The role requires experience deploying LLM solutions in production and integrating AI models into workflows. Suggested revision: Diagram a research workflow from user request through model call, output review and delivery; explain where reliability, latency and cost trade-offs arise.
Agentic frameworks and MCP
Posting: Agentic-framework experience and working knowledge of MCP for tool and data integration are required. Suggested revision: Sketch an agent that retrieves data through a tool, decides its next step and returns a sourced result; explain what the framework and MCP each contribute.
Statistics and machine learning for screening
Posting: Strong grounding in statistics and machine learning is required, and the fellow may build screening tools. Suggested revision: Explain how you would test a hypothetical screening model, including a baseline, evaluation metric and risk of misleading results.
Investment-workflow translation
Posting: Fellows work with portfolio managers and analysts to turn ambiguous investment workflows into technical solutions. Suggested revision: Take a hypothetical research request and write clarifying questions, a narrowly scoped first deliverable and the trade-offs you would explain to its users.

Skills to revise

Python data pipelines
Posting: High Python proficiency is required, and building data pipelines and research automation is a responsibility. Suggested exercise: Use sample data to build a Python pipeline that validates inputs, transforms records and produces a reproducible research summary.
Build an agentic workflow
Posting: The role requires LLM applications built with agentic frameworks; LangGraph, OpenAI Agents SDK and Claude Agent SDK are named examples. Suggested exercise: In one of those frameworks, prototype a tool-using workflow and demonstrate how it handles a failed tool call.
Connect models and data tools
Posting: Familiarity with OpenAI, Anthropic or Google model-provider APIs and working knowledge of MCP are required. Suggested exercise: Prototype a model call with a data-tool connection, then explain the input, output and integration boundary. MCP can be the tool interface for this exercise.
Use agentic coding tools
Posting: Fluency with tools such as Claude Code and Codex is required. Suggested exercise: Use one to implement a small Python change, inspect its edits and tests, and explain what you verified yourself before accepting the result.
Scope and present a desk-facing solution
Posting: The fellow must work directly with non-technical stakeholders, explain trade-offs and ship amid ambiguity. Suggested exercise: Role-play a request for faster research screening; ask clarifying questions, propose a small deliverable and present its limitations and deadline trade-offs in plain language.

Your preparation checklist

Prepare a production LLM walkthrough
The role calls for user-facing experience scoping, building and deploying LLM solutions. If you have a relevant project, prepare a brief walkthrough of the user need, your contribution, deployment choices, results and lessons learned.
Sketch an investment-research workflow
As a preparation exercise, diagram how an agentic workflow could support a portfolio manager’s research or screening. Show data inputs, tool access, human review, failure handling and how you would measure usefulness; label it as a proposed design, not BAM’s process.
Rehearse an ambiguous stakeholder conversation
Prepare questions for a portfolio manager or analyst about their current workflow, constraints and decision points. Practice turning the answers into a scoped first deliverable and explaining technical trade-offs in plain language.
Assemble technical evidence
Identify examples, if available, that demonstrate Python engineering, statistical or machine-learning reasoning, production model integration and use of an agentic framework. Be ready to explain API choices, MCP-based tool or data integration, and how you used agentic coding tools.
Plan a pilot and evaluation
The posting emphasizes testing emerging LLM approaches on live investment problems. Outline how you would compare an approach with an existing workflow, check output quality and reliability, and decide whether it is ready for broader use.
Prepare for the rotation
Map out how you would learn an unfamiliar in-house AI platform while contributing production tooling, then adapt to different investment teams. Use a past example, if available, to illustrate independent work across stakeholders and deadlines.
Bring targeted questions
Ask how the central AI team hands off work to investment desks, how PMs and analysts define a useful tool, what production standards apply, and how the fellowship identifies a potential permanent desk placement.

Availability last checked: Nov 4, 2026. Status can change after a check; confirm on the employer’s website. Report an issue · Our checking policy

KEEP EXPLORING

Similar AI jobs

Related skills and specializations in United States. Matched to this role, not your profile.

Explore more jobs

Finding similar opportunities…