Open to opportunities

Shagun Paul

@shagunpaul

Message

AI/ML engineer specializing in production LLM systems, evaluation, and deployment.

India

Message

What I'm looking for

I seek roles where I can productionize LLM/NLP systems, build evaluation and monitoring pipelines, work on scalable ML infrastructure, and collaborate in cross-functional teams focused on measurable impact.

I am an AI/ML engineer with 4+ years building production-grade systems across NLP, computer vision, and speech analytics. I design and deploy LLM-driven solutions, NL2SQL workflows, and robust evaluation frameworks that improve model reliability and reduce failure modes.

My work includes architecting semantic validation pipelines, training BERT-based multi-class failure classifiers, and deriving evaluation datasets from live traffic to enable continuous benchmarking. I have built multilingual OCR/NLP pipelines, real-time speaker recognition at scale, and geospatial object detection models with strong precision.

I thrive in cross-functional teams to productionize models on Azure, containerize services with Docker, and implement CI/CD for reproducible ML workflows. I focus on measurable impact—reducing manual analysis time, improving inference reliability, and optimizing system performance for enterprise deployments.

Experience

Work history, roles, and key accomplishments

Current

AI/ML Engineer

Current

XansrMedia

Jan 2026 - Present (2 months)

Architected and productionized an LLM evaluation framework for domain QnA systems with semantic similarity scoring and automated failure tagging; trained a BERT-based multi-class failure classifier and implemented continuous benchmarking pipelines to improve answer reliability in high-traffic scenarios.

LLM Prompt Engineering BERT Semantic Similarity Scores NL2SQL Azure Docker

AI/ML Consultant

Government of India

Mar 2023 - Oct 2025 (2 years 7 months)

Led cross-functional AI initiatives across document intelligence, vision analytics, and speech processing, reducing manual analysis time by over 60% and delivering a speaker recognition system with 95% accuracy across 10K+ hours of audio.

Multilingual OCR Speech Recognition Computer Vision Apache Spark MLFlow Docker NER

Data Scientist

Skylon International LLC

Aug 2022 - Mar 2023 (7 months)

Developed ARIMA and LSTM forecasting and anomaly detection models and built interactive Streamlit dashboards, contributing to a 15% reduction in inventory overstock variance.

ARIMA Anomaly Detection Streamlit Plotly Python Data Visualization

Software Engineer

CloudData Technology LLC

Feb 2020 - Aug 2021 (1 year 6 months)

Developed high-throughput REST APIs in Python/Flask and optimized MongoDB schemas, improving query performance by 60%, and integrated AI-driven modules into production systems.

Python Flask REST APIs MongoDB Performance Optimization Docker CI CD AI Integration