Trillium is a leading proprietary trading firm active in US Equities, US Options, Canadian Equities, and OTC Equities. With trading technology built, tested, and optimized in-house, our engineers and technologists collaborate with our traders to deliver tools and functionality designed to support their needs. Our passion for the markets and our drive to succeed have propelled our track record of growth and profitability for over 20 years. At Trillium, we are a team of problem solvers and independent thinkers dedicated to seizing market opportunities.

We are headquartered in New York City, with additional trading floors in Chicago and Miami. To learn more about Trillium, visit www.trlm.com

The Role

Trillium has traded its own capital for over twenty years. Over that time we’ve built an unusually complete dataset — every trade and every trader’s full history, from first day onward. This role exists to put that data to work: to uncover what it can tell us about trading, risk, and the business, and to build the data science function that will keep doing so.

As our first dedicated Data Scientist, you’ll own the function from the start: building the models and reporting that answer the firm’s most important questions, and putting results in front of leadership. The first priorities are already defined; over time, we expect you to become a partner in setting new priorities and to open up lines of inquiry we haven’t thought of yet. The problems are concrete. Which early indicators separate traders who go on to sustained success from those who don’t. How to measure and manage risk as we expand into options, futures, and prediction markets. How to forecast headcount and seat demand across three offices. How to make a growing body of unstructured text useful.

You’ll report to our Managing Director of Trading and work directly with traders, technologists, and business leadership. A data engineering partner keeps pipelines and datasets clean and reliable so your time goes to analysis and modeling. There is no analytics organization yet; you’ll help define what it becomes. We’re looking for someone who is rigorous, self-directed, and comfortable being the only expert in the room — and who uses AI tooling as a normal part of how analytical work gets done.

Responsibilities

Build and maintain the dashboards and reporting that leadership, traders, and business teams use to track trading performance, risk, headcount, and firm-level metrics
Lead deep-dive analyses and predictive modeling on the firm’s strategic questions, owning each from framing through delivery
Apply NLP to unstructured text across the business — internal communications, survey and feedback data, and external text sources — building classifiers, extracting themes, and summarizing at scale
Set research priorities directly with leadership and keep them aligned with firm strategy
Work with your data engineering partner to make new datasets fast, clean, and reliable, and with technology and business teams to scope requirements and expand what the firm can measure
Bring automation and forward-looking analysis into the firm’s daily operating rhythm
Make AI and LLM tooling a standard part of the workflow — exploratory analysis, code and query generation, synthesis of unstructured data — and help set how the firm uses these tools responsibly

Requirements

Strong analytical judgment: able to frame an ambiguous problem, choose an appropriate method, validate the result, and state clearly what a conclusion does and does not support
Master’s degree or higher in Data Science, Statistics, Economics, Mathematics, Computer Science, or a related field, or equivalent demonstrated experience
4+ years in data science, analytics, or business intelligence, including independent ownership of modeling work end to end
Strong Python (required) for analysis, statistical modeling, and visualization
Proficiency with SQL (e.g., PostgreSQL, MySQL) for querying and manipulating data
Hands-on experience (required) applying NLP and LLM-based methods to unstructured text at scale, including evaluating output quality and judging which approach fits a given problem