Location: New York, NY, United States

FMX is seeking a Senior AI Research Engineer to help design, build, and scale Skynapse, FMX's agentic AI framework for rates and derivatives exchange workflows.

Skynapse uses a central orchestrator to understand requests, decompose work, route tasks to specialized Subject Matter Expertise agents, invoke tools, retrieve enterprise data, apply FMX domain knowledge, and synthesize responses, reports, workflow outputs, or system actions.

The role combines production engineering, applied LLM research, agentic workflow design, enterprise integration, and future work with open-weight models that may require evaluation, adaptation, fine-tuning, governed deployment, and inference optimization.

Responsibilities

• Design, build, evaluate, and maintain production-grade AI applications for FMX's rates and derivatives exchange business.

• Develop the Skynapse orchestration layer, including task decomposition, agent routing, tool coordination, context management, response synthesis, and execution monitoring.

• Apply model-level LLM knowledge to improve reasoning quality, retrieval quality, agent reliability, latency, cost, safety, and user experience.

• Evaluate when to use commercial LLM APIs, open-weight models, fine-tuned models, embedding models, rerankers, smaller specialized models, or deterministic software.

• Build agentic workflows using retrieval-augmented generation, semantic search, structured outputs, function/tool calling, planning, workflow orchestration, and human review.

• Integrate agents with FMX enterprise data sources, internal APIs, market data systems, reference data, documents, search indexes, and code repositories.

• Design guardrails for permissions, audit logging, approval workflows, escalation paths, fallback behavior, tool-use limits, source attribution, and production kill switches.

• Create benchmark datasets, regression tests, red-team scenarios, human review workflows, and production monitoring for LLM and agentic systems.

• Mentor FMX engineers in LLM architecture, model behavior, agent design, evaluation, retrieval, tool integration, and production AI engineering.

Qualifications

• Bachelor's degree in computer science, machine learning, AI, mathematics, engineering, statistics, computational linguistics, or a related technical field.

• 5+ years of professional software engineering experience, including production system design, deployment, and support.

• 3+ years of hands-on experience with LLMs, deep learning, NLP, or advanced AI systems.

• Strong academic or research background in machine learning, deep learning, NLP, transformers, LLMs, generative AI, model evaluation, or related areas.

• Demonstrated understanding of transformers, attention mechanisms, tokenization, embeddings, pretraining, instruction tuning, fine-tuning, alignment, inference, context windows, decoding strategies, and evaluation.

• Experience building LLM applications beyond basic prompting, including RAG, structured outputs, function/tool calling, agent orchestration, evaluation, and production monitoring.

• Strong Python skills and experience writing clean, tested, maintainable, production-quality code.

• Experience evaluating, adapting, fine-tuning, or deploying open-source or open-weight language models.

• Practical understanding of LLM and agent failure modes, including hallucination, prompt injection, retrieval errors, tool misuse, reasoning errors, data leakage, unsafe automation, and non-deterministic behavior.

• Strong understanding of enterprise security, privacy, access control, entitlementing, auditability, and responsible AI considerations.

• Ability to communicate complex AI concepts clearly and drive projects from concept through production deployment.

• Strongly Preferred Qualifications

• Master's degree or PhD in computer science, machine learning, AI, NLP, statistics, mathematics, engineering, or a related field.

• Research experience or publications in LLMs, transformers, NLP, deep learning, retrieval, alignment, inference optimization, model evaluation, or agentic AI systems.

• Hands-on experience with supervised fine-tuning, instruction tuning, LoRA, QLoRA, parameter-efficient fine-tuning, preference optimization, distillation, quantization, or domain adaptation.

• Experience with model serving and inference optimization tools such as vLLM, TensorRT-LLM, Hugging Face TGI, ONNX, batching, caching, GPU utilization, and latency/cost optimization.

• Experience selecting and evaluating open-weight models for enterprise use, including trade-offs across model size, latency, quality, context length, licensing, hosting, security, cost, and governance.

• Experience applying AI in financial markets, exchanges, trading platforms, rates, derivatives, market data, risk, surveillance, clearing, or client onboarding.

• Hands-on experience with LangGraph, LangChain, LlamaIndex, Semantic Kernel, AutoGen, CrewAI, or similar orchestration tools.

• Experience with vector databases, hybrid search, semantic retrieval, reranking, OpenSearch, Elasticsearch, pgvector, FAISS, Pinecone, Weaviate, or Milvus.

• Experience with kdb+/q, time-series databases, order book data, trade data, mark-outs, liquidity analytics, or trading behavior analysis.

• Experience with AI observability, LLMOps, model monitoring, prompt/version management, evaluation dashboards, governance, and enterprise controls.

• Experience with C++ or other systems programming languages is a plus, especially in exchange, trading, market data, or high-performance systems environments.

 

Compensation Expectations: $400,000  - $500,000 Total

 

#LI-JM3

This is a model-aware AI research engineering role. FMX is looking for a senior engineer who can build production systems and also understand LLMs below the application layer - transformers, training and fine-tuning methods including those with open weights, inference behavior, evaluation, and failure modes. The role is not aimed at candidates whose AI experience is limited to API wrappers or traditional ML without substantial LLM and agentic AI depth.