Skip to main content

Irys for Word is here. Install Free →

DE

Devansh

Head of AI at Irys, machine learning engineer, AI researcher, and writer

LinkedIn →X →Artificial Intelligence Made Simple→irys.ai

Overview

Devansh is the Head of AI at Irys, the AI-native legal workspace built by Iqidis, Inc. A machine learning engineer and AI researcher, he writes the widely read newsletter Artificial Intelligence Made Simple and has built machine learning systems across legal technology, finance, and public health, with research in deepfake detection and retrieval-augmented generation. At Irys he leads the AI behind matter-aware retrieval, citation grounding, and legal reasoning.

About

Devansh is a machine learning engineer and AI researcher who has built systems across finance, public health, e-commerce, and legal technology. He studied computational mathematics at the Rochester Institute of Technology, where he also contributed to academic research on deepfake detection. He has worked as an AI consultant and was a co-founder and head of data science at an early-stage startup, and his work spans retrieval-augmented generation, model evaluation, and reducing hallucinations in production systems.

As Head of AI at Irys, he leads the work on the problems that make legal AI hard: matter-aware retrieval, separating legal retrieval from legal reasoning, grounding citations in real authority, and treating hallucination as several distinct problems rather than one. The goal is an AI a lawyer can trust to cite what is real and reason about what matters.

He is also the writer behind Artificial Intelligence Made Simple, a newsletter that explains the most important ideas in AI to a technical audience and has been read in more than 150 countries. That same plain, rigorous voice runs through the technical writing he publishes for Irys.

At a glance

Head of AI

at Irys (Iqidis, Inc.)

30K+

newsletter subscribers

150+

countries reached

Legal AI + ML

focus areas

Areas of expertise

Machine learningAI researchLarge language modelsRetrieval-augmented generationLegal AIHallucination mitigationCitation groundingData scienceAI evaluationTechnical writing
“Legal AI fails in more than one way, and most tools only fix one of them. The interesting work is separating retrieval from reasoning, so the model cites what is real and reasons about what matters.”
Devansh

Published by Devansh

Why Legal AI Hallucinations Are Three Different Problems, and Most Tools Only Catch OneThe Real Cost of Running AIAI Market Report Feb 2026: Ten Frontier Models in 28 DaysHow to Use AI for Cutting-Edge Research

Frequently asked

About Devansh

Devansh is the Head of AI at Irys, the AI-native legal workspace built by Iqidis, Inc. He is a machine learning engineer and AI researcher and the writer behind the newsletter Artificial Intelligence Made Simple.

He is known for the newsletter Artificial Intelligence Made Simple, read in more than 150 countries, and for his work building production machine learning systems, including research on deepfake detection and retrieval-augmented generation. At Irys he leads the AI behind matter-aware retrieval and citation grounding.

As Head of AI he leads the technical work behind the platform: matter-aware retrieval, separating legal retrieval from legal reasoning, grounding citations in real authority, and reducing hallucinations so the output is something a lawyer can rely on and verify.

It is the newsletter Devansh writes, explaining the most important ideas in AI to a technical audience of engineers, researchers, and operators. It has reached readers in more than 150 countries.

He has built machine learning systems across finance, public health, e-commerce, and legal technology, worked as an AI consultant and a head of data science, and contributed to academic research on deepfake detection. His focus areas include retrieval-augmented generation, model evaluation, and hallucination mitigation.

Author page · Head of AI at Irys, machine learning engineer, AI researcher, and writer · Last updated June 2026.

All Irys authors →

See the legal AI Irys is building

Irys grounds its output in verifiable authority, keeps client data out of training, and is built by lawyers and AI researchers. See how it fits your practice.

Book a Demo