Machine Learning Researcher

Graviton is a privately funded quantitative trading firm in Gurugram that trades across asset classes and venues using time series analysis, filtering, classification, stochastic models and pattern recognition over terabytes of market data. This seat applies machine learning, specifically deep learning, to trading problems. You would improve tick by tick order book feature sets with modern preprocessing, develop new deep learning models over the firm's in house feature pool and compute infrastructure, and build a scalable pipeline for producing predictive models across global markets. Finding new sources of predictive signal, proving they improve the existing models, and getting them into the strategy development pipeline is the core measure of the work. Implementation happens in Python and C++ alongside quant researchers and developers. The experience floor is genuinely low at two years, but the academic bar is not: a Masters or PhD is required.

What the posting requires:
- A Masters or PhD in Computer Science, Mathematics, Statistics or a related field. This is listed as a qualification, not a preference.
- At least two years of demonstrated ML or AI research experience, either in a professional setting or at a reputable academic institution. Academic research counts here, which is unusual and worth noting.
- Software engineering experience in Python and C++.
- Experience with TensorFlow, Keras or PyTorch is described as highly desirable.
- A track record of academic publications is preferred, not required.

What the work actually looks like:
- Lead research applying machine learning across a wide variety of datasets and trading problems.
- Track academic research and bring techniques from other fields into trading problems.
- Improve tick by tick order book time series feature sets using current preprocessing methods.
- Build and extend deep learning models against a large in house feature pool and compute cluster.
- Build a scalable pipeline for producing predictive models across global markets.
- Discover new sources of predictive signal, verify they improve existing models, and integrate them into the firm's strategy pipeline.
- Work with quant researchers and software developers to move research into production in Python and C++.
- Advise the infrastructure team on hardware and software choices for ML research.

Location and working pattern: Gurugram. The posting describes cubicle free offices and a non hierarchical culture, which implies an in office team, though it does not state a fixed number of office days.

Benefits stated in the posting: competitive compensation, an annual international team outing, fully covered commuting expenses, health insurance, catered breakfast and lunch, gym and sports club memberships, and four weeks of annual leave in addition to market holidays.

Honest fit guidance: this is a research seat at a proprietary trading firm, so the work is model quality and signal discovery rather than product engineering. If you have a strong research record and two years of serious ML work, the low year count will not hold you back here. If you do not have a Masters or PhD, this specific posting is not the one to spend an application on. No interview process is published.