The company builds an AI-powered product network that helps governments and businesses keep trade moving while making the global economy more resilient and secure.
This is a senior staff role leading the AI behind the company Nexus, an agentic platform for global trade.
The work centers on a constant flow of public and proprietary trade data, including shipments, customs filings and trade documents, turned into real-time risk, compliance and enforcement decisions for governments and large companies.
Responsibilities
Set the technical direction for agentic systems, including the architecture and roadmap for models and agents that process trade documents such as shipments, customs declarations, invoices and bills of lading as they arrive, and turn them into risk, classification and compliance decisions in real time.
Lead the design of dependable, cost-efficient systems covering ingestion, document understanding, entity resolution, inference and model routing at global trade scale.
Fine-tune and post-train models (SFT, DPO, RL with domain graders) using analyst feedback, production results and agent traces, but only where it clearly outperforms prompting, and keep models improving over time.
Build the evals, tracing and feedback loops that decide what gets released and catch production regressions.
Mentor senior engineers, raise the technical bar across Engineering, and help shape hiring and leveling for ML talent.
Work with engineers, data scientists, product managers and customers to propose new business ideas and turn unclear problems into shipped capabilities.
Requirements
BS or MS in Computer Science, Machine Learning or a related field, or equivalent experience.
8+ years building and shipping production ML systems, with a record of product-oriented delivery, including recent work on LLM-powered products.
Deep expertise in PyTorch and the modern Python ML stack, such as Hugging Face, scikit-learn, NumPy and Pandas or Polars.
Hands-on experience building production LLM applications, including retrieval-augmented generation (RAG), tool and function calling, structured outputs and agentic workflows.
Experience shipping agentic systems and handling their failure modes: tool use, planning, state, recovery and human-in-the-loop.
Experience post-training or adapting LLMs (SFT, DPO, RLHF/RLAIF, reward modeling) for a production use case.
Strong evaluation practice, including building eval suites, golden datasets and LLM-as-judge pipelines.
Good judgment on when to use a frontier model API, a fine-tuned open-weight model or a classical ML model, weighing cost, latency and accuracy.
Drive to apply state-of-the-art techniques to real problems, and willingness to work hands-on with the data and domain to reach measurable results.
Strong commitment to engineering quality, supportability and knowledge sharing.
Strong written and verbal communication.
Daily use of AI coding tools such as Cursor and Claude Code.
Preferred: experience building multi-agent systems and agent tooling, such as LangGraph.
Preferred: hybrid retrieval with vector search, re-rankers and embedding fine-tuning.
Preferred: vision-language models for document understanding of invoices, bills of lading and customs forms.
Preferred: multilingual NLP.
Preferred: LLMOps with MLflow, Databricks Mosaic AI and guardrails.
Preferred: safety, guardrails or auditability for AI in regulated or high-stakes areas such as customs, legal, healthcare or finance.
Technologies used: Python and Spark; Docker, Git, Kubernetes, Swagger/OpenAPI, AWS and Azure; Databricks, OpenSearch and Postgres; PyTorch, Hugging Face, FAISS and MLflow; vLLM, LangGraph, LangSmith or Braintrust, and frontier model APIs.
Conditions
The role is based in Brooklyn, NY, USA, with hybrid flexibility. Candidates must live within commuting distance of the Williamsburg, Brooklyn office.
Base salary range: $205,000 to $240,000 USD per year. Final pay depends on factors such as experience, education, critical skills and business needs.
Most full-time offers include equity, which is part of total compensation. The role also includes either a discretionary bonus or a variable compensation plan.
The recruiter can share specific pay and benefit details during the hiring process.
Offers come from an the company's site email address. The company never asks for payment during interviews or onboarding.
Benefits
Flexible time off to manage your own schedule and balance.
Paid parental leave: 14 weeks for non-birthing, adoptive and foster parents, and up to 26 weeks for birthing parents, all paid at 100% of base salary.
Medical, vision and dental coverage with generous employer contributions. The high deductible health plan is 100% employer paid, with an employer contribution to your HSA and an optional FSA.
Life, short-term and long-term disability, and AD&D insurance at no cost to you.
Fidelity 401(k) retirement savings.
Pre-tax commuter benefits for transit or parking.
Free premium Calm subscription for meditation, sleep and mindfulness.
Pet insurance through Wishbone and a Total Pet vet and telehealth discount plan.