Looking for a job

Ishan Singh

@ishansingh

Message

Machine learning engineer specializing in production ML systems, multimodal models, and ROI-driven solutions.

India

Message

What I'm looking for

I seek roles where I can combine research and production ML—multimodal models, scalable pipelines, and optimization-focused projects that drive measurable product and revenue impact.

I am a machine learning engineer with a strong foundation in data science and mathematics from Chennai Mathematical Institute and Indian Statistical Institute. I design and deploy production ML/DL systems, having built 12 models serving 100K+ monthly users across 50+ football leagues and delivering measurable business impact.

My work spans multimodal research and production: I reduced multimodal LLM hallucinations by 30% through semantic-preserving audio compression and improved prediction accuracy from 60% to 82% at VUSport using GNNs and ensemble methods. I have experience with end-to-end pipelines, model deployment, optimization, and uncertainty-aware portfolio systems that generated 18% sustained monthly ROI.

I focus on rigorous evaluation, scalable engineering, and actionable business outcomes—combining signal processing, deep learning, and optimization to improve model reliability, user engagement, and revenue. I enjoy tackling hard ML research problems and translating them into robust production solutions.

Experience

Work history, roles, and key accomplishments

Current

Summer Research Fellow

Current

EleutherAI

Jul 2025 - Present (11 months)

Selected from 500+ applicants for EleutherAI SOAR to reduce multimodal LLM hallucinations via semantic-preserving audio compression, achieving a 30% reduction on benchmarks and processing 500GB of audio with 1,500 GPU-hours.

PyTorch Transformers Audio Processing Semantic Compression Model Training

Machine Learning Engineer

VUSport

May 2023 - Mar 2025 (1 year 10 months)

Built and deployed 12 production ML/DL models serving 100K+ monthly users across 50+ football leagues, raised prediction accuracy from 60% to 82% and delivered a 28% engagement increase and 15% churn reduction.

CatBoost XGBoost GNN MLFlow fastAPI

Data Science Intern

VUSport

May 2022 - Jul 2022 (2 months)

Reduced feature space by 40% via Random Forest feature selection and built an LSTM that improved time-series prediction accuracy by 15%, informing production model strategy and reducing training time by ~30%.

Random Forest Time Series Model Evaluation Python scikit learn

Research Scholar

IIT Delhi

May 2019 - Jul 2019 (2 months)

Developed a Bayesian ML model that improved cyclone intensity RMSE by 12% (18.5 → 16.3 knots) using satellite data analysis across 73 cyclones and large-scale geospatial datasets.

Time Series Satellite Data Processing RMSE Reduction Python Statistical Analysis

Education

Degrees, certifications, and relevant coursework

Chennai Mathematical Institute

Master of Science, Data Science

2021 - 2023

Completed an M.S. in Data Science with coursework and projects in machine learning, statistics, and large-scale data processing from Aug 2021 to May 2023.