Artificial Intelligence Analyst

As an Artificial Intelligence Analyst at CryptoMize, you apply machine learning to live intelligence problems with data unlike textbook corpora: multilingual sentiment streams, electoral histories across 18 countries, media-mention floods and coordinated-behavior traces. You own models end-to-end, framing questions, engineering features, validating honestly, and explaining results to strategists who act on them. This is a full-time, permanent position based at our New Delhi HQ with immediate joining available.

What you'll do

• Build and run learning models on engagement corpora including sentiment trajectories, movement prediction, anomaly and coordination detection
• Frame engagement questions as learnable problems: define what to predict, from what signal, and validate against what ground truth
• Engineer features from messy multilingual media, social and survey data to turn raw input into decision tools
• Validate honestly through backtesting, leakage control, and communicating uncertainty instead of hiding it
• Support the practice's electoral forecasting work, models whose historical accuracy across countries is the product's signature
• Write model documentation strategists actually read: what it predicts, what it cannot, and how confident to be
• Partner with platform teams to move validated models into production pipelines and REST interfaces

What you bring

• Machine learning fundamentals: regression, classification, ensembles, and judgment about when simple beats clever
• Deep learning literacy with sequence and transformer architectures applied to text and time-series
• NLP experience on noisy multilingual corpora including tokenization, embeddings and transfer learning
• Feature engineering from media, social and behavioral data
• Rigorous validation practices: backtesting design, leakage control and honest baselines
• Python for ML with pandas, scikit-learn and PyTorch or TensorFlow
• SQL on multi-gigabyte datasets
• Ability to communicate uncertainty, confidence, error and limitations to non-specialists
• Experiment discipline: hypotheses logged, results reproducible
• Discretion with client datasets (NDA-grade confidentiality)
• Domain curiosity across political, reputation and security questions

Nice to have

• Python 3.11 proficiency
• Jupyter-based analysis repositories with Git review discipline
• PostgreSQL for feature and training data
• Experiment tracking tools like MLflow

What we offer

• Compensation discussed during screening, determined by depth of demonstrated skill in the role, classification and scope of client engagement, urgency of active project requirements, and track record across CryptoMize engagements
• Full-time permanent position
• Based at New Delhi HQ

About CryptoMize

CryptoMize is a practice applying machine learning to intelligence problems, with a forecasting record staked on accuracy across electoral and engagement work spanning 18 countries.