AI/ML Research Intern
Trading Systems
Summer 2026
What if you could teach a machine to think like a professional futures trader?
Most internships have you updating spreadsheets and sitting in meetings. This one has you building a machine learning system that makes real trading decisions on live NQ futures markets — and learning from two of the sharpest systematic traders working today.
Trifecta Consulting Group is a Berkshire-based technology firm building an AI-powered futures signal engine. Our ML model watches the market, detects institutional-grade setups, and decides when to trade — autonomously. We're now entering Phase 3: training the machine to learn from expert human judgment.
That's where you come in.
The Project
Coleman s a live trading system running on TradeStation SIM. It processes real-time futures data through a multi-layer signal engine: Multiple Time Frame “MTF” trend detection, session level proximity, order block confluence, Goldbach dealing range analysis, gap bias scoring, and time macro weighting. Every trade decision is logged with 50+ decision variables in a Supabase database.
We have two expert traders who have been trading futures professionally for 20+ years. They use a 1-to-20 conviction scoring system to evaluate setups.
Phase 3 is the Machine Learning “ML” training phase. We will have labeled trade data, a structured feature set, and two domain experts willing to narrate their decision-making in real time. Your job is to turn that into a model that makes expert-level decisions.
What You'll Actually Build
- A feature engineering pipeline that transforms raw trade audit data into ML-ready inputs
- An XGBoost / LightGBM classifier trained on expert level labeled trade decisions
- A back testing framework that validates Coleman's signal accuracy against historical outcomes
- Statistical analysis of which signals actually predict winning trades vs which are noise
- The groundwork for a Phase 4 computer vision model trained on labeled chart screenshots
- A model evaluation dashboard showing signal importance, accuracy curves, and live performance
By the end of the summer you will have built a production-grade ML system that is actively influencing real trading decisions. That is not something most college graduates can say.
The Tech Stack
You'll work with tools that are standard in quantitative finance and ML research:
- Python — pandas, numpy, scikit-learn, XGBoost, LightGBM, matplotlib
- Supabase (PostgreSQL) — querying 50+ column trade decision tables
- Jupyter / VS Code — model development and analysis
- GitHub — version controlled research notebooks and pipeline code
- Lovable / React — optional frontend work if you want to build visualizations
- TradingView Pine Script — reading and understanding existing indicator logic
No prior finance or trading knowledge required. We will teach you what you need to know about futures, ICT methodology, and systematic trading. What we need from you is mathematical rigor and the ability to think clearly about data.
What You'll Learn
Quantitative Finance
- How institutional futures markets actually work — order blocks, liquidity sweeps, dealing ranges
- How professional traders quantify conviction and structure position sizing
- How systematic trading systems are designed, validated, and deployed
- The difference between a signal that backtests well and one that actually works live
Machine Learning in Production
- Feature engineering for time series financial data
- Supervised classification on imbalanced real-world datasets
- Model validation techniques that prevent overfitting on financial data
- Bridging expert human judgment and algorithmic decision-making
- The architecture of dual-input models combining structured features and computer vision
Research and Communication
- How to work directly with domain experts to extract tacit knowledge into structured features
- How to present model findings to non-technical stakeholders
- How to document research so it can be reproduced and extended
Who We're Looking For
You don't need to know anything about futures trading. You do need to be the kind of person who gets genuinely excited about hard problems and follows through on them.
Required
- Sophomore, Junior or senior at Williams College, MCLA, or comparable institution
- Strong coursework in mathematics — linear algebra, probability, statistics
- Programming proficiency — you can write clean, documented code without being hand-held
- Comfort with ambiguity — this is research, not a problem set with a known answer
- Intellectual curiosity — you read things that weren't assigned and ask questions that weren't on the syllabus
Strong Plus
- Machine learning coursework — even intro level (CS 381 or equivalent)
- Experience with pandas and data manipulation
- Familiarity with SQL or willingness to learn quickly
- Interest in financial markets — you don't need to trade, but you should find markets interesting
- Statistics coursework beyond intro — regression, hypothesis testing, experimental design
Not Required
- Prior finance or trading internship experience
- Knowledge of futures markets or trading methodology
- Full-stack development skills
- A GPA above 3.5 — we care about how you think, not your transcript
The Opportunity
This is not a typical summer internship. You won't be supporting someone else's project. You will own a significant piece of a live, production AI system that is actively trading in real markets.
You'll work directly with the founders and have regular access to the domain experts whose judgment the ML model is learning from. You'll see how a fintech product gets built from first principles — the methodology extraction, the rules engine, the signal architecture, the training data pipeline, and the model deployment.
If you're considering a career in quantitative finance, data science, or AI research, this internship gives you a portfolio project that is genuinely difficult, genuinely interesting, and genuinely differentiated from what most candidates have.
Compensation and Structure
Location and hours: Can be remote or in person in Williamstown, MA. You can work at your own pace as long as defined targets and milestones are met.
Compensation: TBD, we can structure an hourly rate or, set up a potential employment opportunity with deferred compensation.
How to Apply
We are not looking for a polished cover letter. We are looking for evidence that you can think clearly and work independently on hard problems.
Send the following to: jfredette@trifectacgrp@gmail.com
Subject line: Coleman ML Intern — [Your Name] — [Your School]
Include:
- A brief paragraph (3-5 sentences) describing the hardest data or math problem you've worked on and what you learned from it
- Your resume or CV — one page preferred
- One example of code you've written that you're proud of — GitHub link, Jupyter notebook, or pasted snippet. It doesn't need to be long. It needs to be clean and well-commented.
- One sentence on why this project specifically interests you
We review applications on a rolling basis. Strong candidates will be contacted for a 30-minute conversation within one week of applying.
Trifecta Consulting Group is an equal opportunity employer. We welcome applications from candidates of all backgrounds.
Williamstown, MA | Summer 2026 | AI/ML Research Intern — Quantitative Trading Systems