Open to opportunities

Mehroz Azyaan

@mehrozazyaan

Message

Senior AI/ML Engineer building scalable LLM and MLOps platforms that deliver production impact.

United States

Message

What I'm looking for

I’m looking to build enterprise AI platforms end-to-end—MLOps, LLM/NLP systems, and AI automation—where I can ship reliable production models, optimize inference costs and latency, and use observability for measurable performance gains.

I’m a Senior AI/ML Engineer with 9+ years building scalable ML platforms, LLM applications, and AI analytics for healthcare, SaaS, and enterprise. I’ve shipped 14 production models serving 2M+ daily predictions, with a focus on reliability, speed, and measurable business outcomes.

At Axonic Technologies (Apr 2022 — Present), I lead a team of 4 engineers operating 14 production AI/ML models across predictive analytics, NLP automation, and intelligent workflows, serving 2M+ predictions per day. I architected an MLOps platform on AWS SageMaker, Kubernetes, MLflow, and Docker—cutting model release cycles from 6 weeks to 48 hours using canary deployment and automated retraining.

I build LLM-powered document intelligence and semantic search using Hugging Face Transformers plus pgvector/Pinecone, processing 40k+ documents per day at 91% retrieval precision. I also engineer real-time GPU inference serving 2M+ predictions daily at p99 latency under 120ms, reducing inference compute cost 38% via quantization and autoscaling.

My engineering ethos is observability and governance: I implemented AI observability with drift detection, feature validation, and SHAP explainability, catching 4 silent regressions before production impact. Previously at Tempus AI and Uptake Technologies, I delivered healthcare ML automation and production predictive systems—designing feature stores, monitoring dashboards, and distributed pipelines while mentoring junior engineers.

Experience

Work history, roles, and key accomplishments

Current

Senior AI / ML Engineer

Current

Axonic Technologies

Apr 2022 - Present (4 years 2 months)

Led a team of 4 engineers operating 14 production AI/ML models across predictive analytics and NLP automation, serving 2M+ predictions per day. Built an AWS SageMaker/Kubernetes/MLflow MLOps platform and LLM semantic search, cutting release cycles from 6 weeks to 48 hours and achieving p99 inference latency under 120ms with 38% lower compute costs.

AWS Sagemaker Kubernetes MLFlow Docker Terraform Canary Deployment Pgvector Pinecone

ML Engineer

Tempus AI

Aug 2019 - Mar 2022 (2 years 7 months)

Built 9 ML pipelines for healthcare analytics, clinical NLP, and predictive modeling across 6 pharmaceutical collaboration studies, automating 75% of manual chart reviews. Trained and deployed models on 18M patient encounter records (up to 0.89 AUC-ROC), and implemented Feast/Delta Lake feature stores and production monitoring achieving 99.9% serving uptime.

PyTorch TensorFlow XGBoost scikit learn ICD 10 Coding Feast Delta Lake SHAP MLFlow

Data Scientist

Uptake Technologies

Jun 2017 - Jul 2019 (2 years 1 month)

Developed predictive analytics and anomaly detection models for IoT across mining, energy, and transportation, processing 500M daily sensor readings. Built time-series failure prediction using isolation forests and autoencoders (92% TPR at <2% FPR) and engineered Spark pipelines that produced 140 features enabling 18-month datasets to train in under 4 hours, while reducing training time by 55%.

Isolation Forest Time Series Anomaly Detection Spark Feature Engineering Model Training Optimization