Peeush Agarwal

Principal MLOps Engineer | AI Infrastructure & LLM Systems

agarwal.peeush@gmail.com
+91 82377 28795
Pune, Maharashtra
LinkedIn

Profile

Principal-level MLOps Engineer with 14+ years of experience designing and operating AI infrastructure at scale. Deep expertise in end-to-end MLOps pipelines on Azure ML, LLM application development (RAG, agentic AI, multi-agent orchestration), and production model deployment. Proven track record leading cross-functional teams, architecting secure ML systems, and driving automation across CI/CD, model serving, and observability layers. Experienced with Kubernetes, MLflow, LangChain, LlamaIndex, and Azure AI Search in production environments.

Experience

Senior ML Engineer

Jan 2025 - Present
The Mosaic Company • Pune, IN

Leading agentic AI and MLOps initiatives with a cross-functional team of 4. Owns end-to-end AI infrastructure including production model deployment on Azure Kubernetes, CI/CD pipelines, model drift monitoring, LLM-based agents, RAG systems, and automated reporting workflows.

  • Productionized 9 Commodity Forecasting models on AzureML Endpoints, deploying to Azure Kubernetes clusters via AzureML Python SDK and CI/CD pipelines; implemented continuous model monitoring using Azure Log Analytics to detect drift and automatically trigger retraining workflows via email alerts.
  • Built an agentic RAG chatbot using Azure AI Search and LlamaIndex, ingesting 8 weekly commodity reports over a 6-month rolling window with an automated ingestion pipeline to keep the vector store current — enabling conversational querying by ELT and Finance teams.
  • Designed and deployed an automated weekly summarization pipeline that applies structured summarization instructions via AzureOpenAI to commodity reports and delivers formatted executive email digests to the ELT team — reducing report turnaround from 2-3 days to 1-2 hours.
  • Architected multi-agent orchestration workflows for planning and execution, enabling autonomous end-to-end task completion and reducing manual intervention.
  • Owned production deployment, monitoring, and safety guardrails for agentic systems, including logging, fallback policies, and human-in-the-loop controls.
  • Mentored cross-functional team of 4 on agent design patterns, experiment-driven development, and KPIs for agent-driven automation.

Senior ML Engineer

Aug 2022 - Jan 2025
Process Point Technologies (PPT) • Pune, IN

PPT is a technical consultant company with expertise in AI, ML, DL, ERP domain. Led engineering initiatives for model deployment and infrastructure.

  • Architected end-to-end MLOps platform on Azure ML including model registry (MLflow), automated retraining pipelines, and CI/CD integration — reducing deployment cycle by 50%.
  • Developed and deployed a Golang-based Traffic Simulator API with optimized model serving, reducing inference time by 40%.
  • Implemented MLOps pipelines for the Mine Vision System, integrating Computer Vision models into AzureML with automated monitoring and feedback loops.
  • Designed and implemented a PowerBI dashboard for microservices observability, providing comprehensive operational insights.
  • Collaborated with cross-functional teams to deliver production-grade ML solutions across computer vision and NLP domains.

Senior ML Engineer

Nov 2021 - Aug 2022
Privacera (Freestone Infotech Pvt. Ltd.)

Privacera is into the data governance domain, based in US.

  • Utilized NER-based spaCy model to identify and mask sensitive entities within textual data, improving pipeline performance by 35%.
  • Built a Medical NER model for entity recognition within clinical text, improving accuracy by 20%.
  • Constructed a feedback-driven pipeline for reassessing model-generated entities based on user feedback, enabling continuous model improvement.

Senior Software Developer

May 2013 - Nov 2021
EuroForm A/S

EuroForm provides Managed Print Services oriented solutions to its client-base worldwide.

  • Supervised end-to-end ML model lifecycle from data collection to deployment and monitoring.
  • Led integration of ML models into cloud-based APIs and production services.
  • Solved highly imbalanced datasets using clustering and targeted feature engineering.

Staff Software Engineer

Jul 2011 - May 2013
Hewlett Packard (HP)

Built REST-based web service and integrated with internal systems to provide data for printer fleet management.

  • Designed and implemented a REST-based web service for printer fleet management.
  • Integrated the web service with internal systems to provide real-time data for monitoring and analytics.
  • Collaborated with cross-functional teams to ensure seamless integration and deployment.

Projects

PhotoMind — Multimodal Personal Photo Organizer (Self-hosted)

Problem: Personal photo libraries are unsearchable without manual tagging. Approach: Built a self-hosted multimodal RAG application using Ollama + Gemma4 for local image captioning and semantic understanding, generating embeddings stored in ChromaDB for vector retrieval. Extracted structured metadata (timestamp, GPS location, category) stored in a SQL database enabling hybrid filtering alongside semantic search. FastAPI backend manages async ingestion jobs triggered by pointing at any local image directory; React + ViteJS frontend enables natural language search and faceted filtering by category, date, and location. Outcome: Fully on-prem, privacy-preserving photo search running on local inference with zero cloud dependency — demonstrating sovereign-data deployment patterns.

Education

Post Graduate Diploma in Data Science in Data Science

2017 - 2018
IIIT Bangalore & upGrad

B. Tech. in Computer Science

2007 - 2011
National Institute of Technology, Durgapur

Skills

Generative AI

AzureOpenAI • Microsoft Agent Framework • LangChain • LlamaIndex • RAG Pipelines • Azure AI Search • Chroma • Agentic AI • Multi-Agent Orchestration

Machine Learning

Deep Learning • NLP • Computer Vision • MLOps • Model Optimization • Feature Engineering • Transfer Learning

Cloud & MLOps

AzureML • MLflow • AWS • GCP • Docker • Kubernetes • CI/CD • Model Registry • Monitoring & Observability

Technologies/Tools

PyTorch • TensorFlow • Kafka • Django • Flask • Jenkins • GitHub Actions • Git

Programming Languages

Python • Golang • C# • SQL

Leadership & Architecture

Technical Leadership • Software Architecture • Distributed Systems • Cross-functional Collaboration • Mentoring & Coaching • Project Roadmaps

Languages

PythonAdvanced
GolangIntermediate

Interests

Currently exploring

LLM agents for logistics automation, RAG pipelines, Real-time inference optimization