About Hytech
Hytech is a leading management consulting firm headquartered in Australia and Singapore, specializing in digital transformation for fintech and financial services companies. We provide comprehensive consulting solutions, as well as middle- and back-office support, to empower our clients with streamlined operations and cutting-edge strategies. Our key clients include top trading platforms and cryptocurrency exchanges.
With a global team of over 2,000 professionals, Hytech has established a strong presence worldwide, with offices in Australia, Singapore, Malaysia, Taiwan, Philippines, Thailand, Morocco, Cyprus, and more.
You'll join our Singapore hub, working closely with Quants, Data Scientists, and Risk Analysis team(RA) on a high-impact production system.
The mission
Drive the ML/DL research agenda for our client-classification engine and adjacent transaction risk control/anti-fraud scenarios across the trading chain. Take the best ideas to production—safely, fast, and cost-effectively. You'll own the end-to-end experiment loop, from hypothesis → backtest → calibration → monitored rollout, so our routing and control decisions improve broker P&L, stability, and abuse prevention.
What You'll Do
Essential (Core)
- Be the key owner for modeling research execution: turn business and user questions into measurable data solutions; collaborate with analysts and engineers to build, test, and iterate.
- Explore and apply ML/DL models: experiment with a variety of algorithms to solve practical integrity and classification problems.
- Tackle related challenges: work on areas like user behavior modeling, fraud/anomaly detection, reputation/risk scoring, or abuse prevention, depending on business needs.
- Own the experiment loop: design fair, time-aware evaluation strategies; conduct robust backtesting and validation; clearly report findings and next steps.
- Partner for production: collaborate with engineers to ship models safely, monitor health, and set up retraining or recalibration triggers.
- Communicate clearly: write concise design docs and experiment summaries that both technical and business stakeholders can act on.
Advanced (Plus)
- Deep learning and time-series: comfortable exploring neural network models (e.g., LSTM, GRU, Transformers) or advanced architectures for temporal/user-sequence data.
- Representation learning: experience with embeddings, autoencoders, or other feature-learning approaches to improve prediction or personalization.
- Working with imperfect data: familiar with handling noisy, incomplete, or partially labeled data (e.g., semi-supervised or weak supervision).
- Experimentation at scale: hands-on with tuning, validation, and reproducibility on large datasets or distributed compute (e.g., Spark/Databricks, MLflow).
- Business/context awareness: able to connect modeling choices to real-world metrics and user/business value.
Minimum qualifications
- 5+ years in Data Science/ML/DL, with a track record of using data and models to solve practical business or user problems.
- Strong Python and SQL; able to write clean, reproducible analysis code in notebooks or scripts.
- Hands-on with at least one ML or DL library (scikit-learn, PyTorch, TensorFlow).
- Solid experiment design fundamentals: careful validation, fair metrics, and an eye for avoiding data pitfalls.
- Effective communicator: able to summarize findings and decisions for both technical and non-technical partners.
Nice to have
- Experience in integrity, fraud, abuse, personalization, recommendations, search, ads, or growth analytics.
- Familiarity with cloud analytics stacks (Spark/Databricks/MLflow) or large-scale experimentation.
- Exposure to classification, anomaly detection, or user segmentation in any domain.
Tech you'll use
Databricks (Spark/Delta/UC/MLflow), Python, Spark SQL, Databricks Workflows/Jobs
