Skip to content
/ LAARI Public

AI-powered HPC platform for financial risk analysis. Run million-scenario simulations in plain English without knowing anything about clusters or parallel computing. Features intelligent workload routing, Gemini-powered reports, and real-time monitoring.

Notifications You must be signed in to change notification settings

riyap10/LAARI

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

LAARI - a HPC Risk Analysis Platform

A platform that makes high-performance computing accessible for financial risk analysis. Analysts can run massive simulations without knowing anything about clusters, schedulers, or parallel computing. Just describe what you want in plain English, and the system does the rest.


Why We Built This

Financial risk simulations often require running millions of scenarios. Traditional HPC setups are powerful but complicated—analysts shouldn't have to worry about cores, nodes, or cloud costs to get results.

Our goal: let users focus on what they want to test, not how to run it. HPC is powerful, but it should be invisible.


Features

Plain-English Job Submission

Type requests like:

"Stress test my portfolio under extreme market volatility."

The system:

  • Figures out the type of analysis
  • Picks parameters like volatility and depth
  • Generates the HPC job automatically

No technical knowledge required.


Smart Workload Routing

Jobs automatically run where it makes sense:

  • Local WebGPU for free, small-scale jobs
  • Cloud HPC clusters for time-critical or high-load jobs
  • Hybrid mode to balance cost and performance

High-risk scenarios get more parallel workers automatically, while routine checks save resources.


Massive-Scale Processing

  • Millions of scenarios in parallel
  • Hundreds of millions of data points
  • Results in seconds instead of minutes or hours

Everything is handled behind the scenes—users just see results.


Executive-Ready Reports

Instead of raw metrics, users get insights they can act on:

  • Risk levels (Critical / High / Moderate / Low)
  • Financial impact in dollars
  • Recommendations for next steps

Reports are HTML-based and presentation-ready. Shareable with stakeholders without extra work.


Persistent Data & Security

All results are stored in MongoDB so you can:

  • Compare current vs past simulations
  • Track portfolio risk trends
  • Keep data secure with Auth0 authentication

Each analyst gets a private workspace where simulation history is protected and accessible only to authorized users.


Real-Time Monitoring

Users can watch:

  • Job progress
  • Resource usage (CPU/GPU/Cloud)
  • Latency and performance metrics
  • Cost breakdowns

Everything updates live, so you know what's happening at all times.


Tech Stack

  • Frontend: React + TypeScript
  • Backend: Node.js + Express
  • Database: MongoDB
  • Authentication: Auth0
  • Compute: Local WebGPU + Cloud HPC
  • AI: Google Gemini API, Groq API

How It Works

  1. User submits a request in natural language.
  2. Gemini interprets it and creates a structured HPC job.
  3. The scheduler routes the job intelligently across compute resources.
  4. The job runs in parallel and results are aggregated.
  5. Gemini generates a plain-language report with actionable insights.
  6. Results are securely stored in MongoDB with Auth0-protected access.

All steps are automated—users never touch the HPC layer.


Why It's Cool

  • HPC without the headache
  • Automatic cost optimization
  • Real-time insights
  • Secure, authenticated access to simulation history
  • Reports that make sense to humans, not machines

Basically, you get all the power of HPC without the learning curve.


Getting Started

# Clone the repo
git clone https://github.com/your-username/hpc-risk-analysis.git

# Install dependencies
npm install

# Start the frontend
npm start

# Start the backend
npm run dev

# Submit a simulation and check results

About

AI-powered HPC platform for financial risk analysis. Run million-scenario simulations in plain English without knowing anything about clusters or parallel computing. Features intelligent workload routing, Gemini-powered reports, and real-time monitoring.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages