Looking for a job

Ishan Goel

@ishangoel

AI/ML Engineer at Turno who raised text-to-SQL benchmark accuracy from 60% to 86% with schema-aware tools and query guardrails.

India

Message

At Turno, I built schema-aware tools and query guardrails for a text-to-SQL system, improving its benchmark accuracy from 60% to 86%. I also developed a scorecard-based LTV classification model and shipped REST APIs for credit risk and business rule models.

At Sprinklr, I fine-tuned ERNIE-Layout for document question answering, reaching 84.1% accuracy on real-world client images. My work has also included a TypeScript SDK for the SUI blockchain, a secure call data analysis platform, and ML pipelines for epigenome segmentation.

Experience

Work history, roles, and key accomplishments

ML Engineer

Turno

Jan 2026 - Sep 2026 (8 months)

Built MCP server exposing read-only Redshift tools to Claude for internal text-to-SQL, improving accuracy from 60% to 86%. Developed and deployed a scorecard-based classification model for Loan-to-Value (LTV) with RESTful APIs, and built a survival model for fraud identification achieving 0.91 AUC-ROC.

MCP Redshift SQL Claude Text To SQL Classification REST APIs CI CD XGBoost Survival Analysis

Product Engineer Intern

Sprinklr

May 2023 - Jul 2023 (2 months)

Designed a multimodal document intelligence solution for visual data extraction, analyzing VQA and key-value models including ERNIE-Layout, LayoutLM, and Geo-LayoutLM. Fine-tuned ERNIE-Layout on DocVQA and Samsung client datasets, achieving 84.1% accuracy, and applied Geo-LayoutLM for key-value extraction with 0.916 F1 for entity labeling.

Multimodal AI Document Intelligence Visual Question Answering ERNIE Layout LayoutLM Geo LayoutLM Fine Tuning Token Classification F1 Score

Software Engineer Intern

Keygen Labs

Oct 2022 - Feb 2023 (4 months)

Developed early versions of Martian's crypto wallet Chrome extension with 500K+ users. Created SUI blockchain SDK in TypeScript and integrated it with the wallet, onboarding 100K+ users, and wrote smart contracts in Move language for NFT rewards on Aptos blockchain.

Chrome extensions TypeScript SUI Blockchain SDK Development Smart Contracts Move Aptos NFT

Machine Learning Intern

Simon Fraser University

May 2021 - Jul 2021 (2 months)

Improved human epigenome segmentation and annotation through Dynamic Bayesian Network (DBN) model (Segway). Debugged, re-architected, and optimized an end-to-end DBN-based ML pipeline on Google Cloud Platform, achieving 89% segmentation accuracy.

Dynamic Bayesian Networks Segway Google Cloud Platform Baum Welch Forward Backward data preprocessing Encode