Architect and optimise infrastructure for training agentic large language models (LLMs) with reinforcement learning (RL), focusing on Mixture-of-Experts (MoE) models above 300 billion parameters.
Responsibilities
Own the architecture of large-scale RL training stack
Profile the full RL loop and cut time per step and GPU idle time
Design asynchronous, fault-tolerant pipelines for model interaction with tools and agent harnesses
Root-cause numerical, distributed and performance issues
Work closely with Research Scientists to test new algorithms and environments
Requirements
MSc or PhD in computer science, machine learning or a related field
Equivalent industry experience
Nice to have: RL post-training frameworks (NeMo-RL, verl, OpenRLHF, TRL)
Nice to have: Training/serving models above 100 billion parameters at 100k+ token contexts
Nice to have: Distributed runtimes and orchestration (Ray, Kubernetes, Slurm) on AWS, Azure or Google Cloud
Build machine learning at UNiDAYS from the ground up, turning business questions into measurable ML problems.
Responsibilities
Find where ML can drive measurable business value and frame it as clear problems with success metrics and a baseline
Source, assess and improve the data needed, closing gaps with product and data teams
Set up processes to clean, label and validate data, including LLM-assisted labelling
Define how success is measured, from offline evaluation to online experiments tied to business outcomes
Shape the training and evaluation harness (experiment tracking, versioning, deployment, monitoring), partnering with the Platform team on the shared capabilities
As an Associate for the Quantitative Trading & Research team, you will primarily focused on supporting Credit & Public Finance Sales. This global sales team distributes corporate & municipal bonds, loans, CLOs, CDS, credit indices, and other corporate-credit related products. You will turn complex data into client-relevant signals, content and trade ideas that support both inbound and outbound client workflows, and own each project end to end: from problem framing and data preparation through methodology, evaluation, production implementation and monitoring.
Compensation
We offer a competitive total rewards package including base salary determined based on the role, experience, skill set and location. Those in eligible roles may receive commission-based pay and/or discretionary incentive compensation, paid in the form of cash and/or forfeitable equity, awarded in recognition of individual achievements and contributions.
Responsibilities
Generate client-tailored marketing materials and analytical content for JPM inventory, with clear source attribution and supporting rationale, sent directly to sales and/or clients, to increase client satisfaction, draw more clients onto our proprietary platforms, and increase client “stickiness”
Develop relative-value and analytics tools that match client interests and holdings with available inventory, generating trade ideas and prioritizing the most actionable opportunities
Build tools and data pipelines that automate or aid sales workflows, combining structured and unstructured market, trading and client data from multiple sources with automated data-quality checks
Design, prototype, build and support AI agents that support sales workflows, for example LLM-based extraction of structured information from unstructured text, with rigorous evaluation against benchmarks and baselines, using error analysis and Sales feedback to improve accuracy
Package analytical methods as reusable tools, skills, plugins and MCP servers that Sales applications and AI agents can call
Requirements
Masters-level or equivalent experience
Strong academic background in a quantitative field (e.g., Mathematics, Physics, Statistics, Economics, Computer Science, Data Science, Engineering, Quantitative Finance)