Overview

QiCAP.Ai is a quantitative investment and trading firm that captures opportunities across asset classes in global markets using high-frequency trading systems and strategies.

Responsibilities

• Design, train and evaluate ML models (gradient-boosted trees, deep sequence models, transformers) on very large volumes of tick-level and order-book data, including online and adaptive approaches (bandits, reinforcement learning) for decisions that must adapt as markets shift.
• Tackle problems where signal is weak and dynamics are only partly observable: market behavioural patterns, cross-correlations across instruments, short-horizon return and volatility forecasting, fill and cancel probability, queue dynamics, and regime and change-point detection.
• Engineer features from raw market microstructure data and build the pipelines that serve them consistently in research and production.
• Make models production-ready under tight latency and memory budgets, using distillation, quantization, pruning and efficient inference.
• Own projects end to end, from hypothesis through modelling, backtesting and simulation to live deployment and drift monitoring, with rigorous validation throughout (walk-forward and purged cross-validation, leakage checks, calibration, honest out-of-sample measurement).
• Direct exposure to production high-frequency trading infrastructure and live markets, where your model's value is measured in PnL, with access to state-of-the-art AI models and large-scale compute.
• Mentorship through code reviews and research discussions, in a collaborative, high-ownership environment where good ideas ship quickly, whoever they come from.

Qualifications

• Currently pursuing, or have completed, a B.Tech in Computer Science and Engineering, Electrical/Electronics Engineering, Data Science, or Mathematics and Computing, with 0-2 years of experience (internships and research work count) and a CGPA of 8.5/10 or above.
• Strong foundations in probability, statistics, linear algebra and optimization (convex optimization, estimation theory or information theory is a plus), and solid ML fundamentals: bias-variance, regularization, overfitting, cross-validation, loss design, and evaluation under class imbalance and distribution shift.
• Hands-on experience with deep learning in PyTorch or TensorFlow, and with classical ML (XGBoost, LightGBM, scikit-learn).
• Proficiency in Python and its scientific stack (NumPy, Pandas, SciPy), along with working knowledge of C++ fundamentals, since models and features must eventually run in low-latency systems.
• First-principles thinking, the ability to take on open-ended problems independently, and a bias for execution: you iterate quickly, measure results honestly, and report negative results as readily as positive ones.
• Time-series modelling (LSTM/TCN, transformers, HMMs, Kalman filters), including on noisy or non-stationary data.
• Reinforcement learning, bandits or online learning; probability calibration and change-point detection.
• Research publications or strong results in Kaggle, quant or ML competitions, or in competitive programming and maths.
• Interest in financial markets, market microstructure and trading.