Back to all work
August 2025 — Present · London

JP Morgan Chase

Staff Software Engineer

Architecting and leading the redesign of the end-of-day risk computation platform processing $14B+ portfolios, while building AI-assisted tooling that accelerates quantitative model migration at scale.

Team6 engineers
DurationOngoing
Stack
Java Kotlin Spring Boot AI/LLM
DomainQuantitative Risk
Projects

What I built here

01

Risk Computation Platform Redesign

Redesigned the end-of-day risk computation platform processing $14B+ portfolios, introducing a model orchestrator engine with dynamic dependency resolution and intelligent workload bucketing.

Challenge

Redundant model execution across portfolios caused excessive compute spend, slow processing times, and frequent SLA breaches during end-of-day risk runs.

Approach

Built a model orchestrator engine with dynamic dependency resolution and intelligent workload bucketing that eliminated redundant model execution across the platform.

Impact

Reduced compute usage by ~35%, processing latency by 30%, and SLA breaches by 80%.

02

AI-Assisted Model Migration Platform

Accelerated the strategic migration of 500+ quantitative risk models by architecting an AI-assisted platform that automated model transformation workflows across QR and engineering teams.

Challenge

Migrating 500+ quantitative models manually required coordinated effort between QR and engineering teams, estimated at 600+ developer days.

Approach

Built an AI-assisted migration platform using LLM integration to automate model transformation workflows, reducing manual intervention across teams.

Impact

Eliminated 600+ developer days of manual effort. Enabled faster model onboarding from 10 days to 5 hours via self-service declarative risk modeling.

03

Platform Reliability Standards

Established platform reliability standards across automated regression testing, execution observability, and operational runbooks for overnight risk computation.

Challenge

Overnight risk computation failures were frequent, with no consistent testing, observability, or incident response patterns across the platform.

Approach

Established automated regression testing, execution observability dashboards, and standardized operational runbooks. Provided technical mentorship to 6 engineers.

Impact

Reduced overnight risk failures by ~45%. Grew engineering capability through mentoring 6 engineers via system design reviews and technical coaching.

Architecture

System design highlights

Portfolio Data
Model Orchestrator
Workload Bucketing
Risk Compute

Key decisions

  • Dynamic dependency resolution to eliminate redundant model execution across portfolios
  • Intelligent workload bucketing to optimize compute distribution
  • Self-service declarative modeling to decouple model onboarding from engineering
  • AI-assisted migration using LLM workflows for model transformation at scale

Trade-offs accepted

  • Dependency graph complexity in exchange for ~35% compute savings
  • AI-assisted migration trades perfect accuracy for massive time savings (600+ dev days)
  • Declarative config adds schema maintenance overhead but cuts onboarding from 10 days to 5 hours