LIVERISK 10,000-path Monte Carlo · 95% VaR / CVaR |CREDITIQ AUC-ROC 0.825 |IRIS LLM ROUTER onboarding latency -70% |SOLARSENSE maintenance cost -35% · output +8% |BILLING FIX ledger error resolved <24h |FINBERT sentiment accuracy 87% |THAPAR IOET · CGPA 8.2 / 10 |LIVERISK 10,000-path Monte Carlo · 95% VaR / CVaR |CREDITIQ AUC-ROC 0.825 |IRIS LLM ROUTER onboarding latency -70% |SOLARSENSE maintenance cost -35% · output +8% |BILLING FIX ledger error resolved <24h |FINBERT sentiment accuracy 87% |THAPAR IOET · CGPA 8.2 / 10 |

AI/ML Engineer — Risk & Quantitative Systems

Parth Deshmukh
builds the models
that price the risk.

I build the systems that sit between raw data and a decision: credit models, portfolio risk engines, and the LLM infrastructure to run them in production. Currently shipping multi-LLM orchestration at an early-stage AI startup, and building the kind of risk & credit tooling I'd want to see inside a markets or risk function at a bank.

View the models View resume

Position SummaryACTIVE

RoleAI/ML Engineer · Iris

FocusLLM Orchestration, Risk Models

EducationB.E. CE, Thapar IET '28

GPA8.2 / 10.00

Core models3 architected

Target functionsRisk · Credit · Quant Eng

LocationBengaluru, IN

Why this page exists

One thesis, three proofs.

Markets need pricing. Lenders need
scoring. Operators need forecasting.

01 — IN PRODUCTION

Experience at Iris

Multi-LLM orchestration, agentic onboarding, and a billing-pipeline fix — live, not simulated.

02 — THE MODELS

LiveRisk, CreditIQ, SolarSense

Monte Carlo risk engines, a credit classifier at 0.825 AUC-ROC, and predictive maintenance.

03 — THE STACK

Tools & Credentials

From Monte Carlo simulation to AWS deployment — plus McKinsey Forward and SEBI certifications.