FinTech · Quantitative ML across Treasury, Credit & Aerospace Research

J.P. Morgan

Role: VP · Quantitative Data Scientist (Liquidity & Credit ML)FinTech · Treasury · Credit Risk · Aerospace ResearchPrior experience — Michele Sanna at J.P. MorganStatus: Completed

Three years of quantitative ML work across J.P. Morgan, spanning treasury, credit risk, and an aerospace research collaboration. Built a streaming intraday-liquidity forecasting engine across CHAPS, Fedwire, TARGET2 and CLS using a Temporal Fusion Transformer with a Bayesian state-space layer for calibrated 15-minute-ahead nostro balances. Designed a corporate credit early-warning system on alternative data (shipping manifests, news NLP, invoice behaviour, satellite indicators) feeding a DeepSurv survival model, with SR 11-7-aligned governance. On a seconded quantitative research collaboration, developed a deep-RL station-keeping policy (PPO with recurrent critic) for autonomous satellite control, with formally-verified safety constraints.

  • Python
  • PyTorch
  • JAX
  • Spark Structured Streaming
  • Kafka
  • Kdb+/q
  • Databricks
  • Delta Lake
  • Snowflake
  • Hugging Face
−22%peak intraday exposure on USD clearing
~$140M/dayof trapped collateral released through smarter payment sequencing
0.89 AUC12-month credit downgrade prediction (vs. 0.71 baseline logistic model)
78%of eventual downgrades flagged ≥90 days in advance
−18%satellite fuel per year vs. baseline MPC controller (aerospace research collab)

Next project

Apple UK (via Deloitte)

Next

Want this kind of work in your business?

Start with the problem, not the platform. Tell us what's slow, manual or breaking.

Send a project briefOr just say hello