Pranay Modukuru

Senior AI Engineer Mar 2025 – Present

• Built a REST API-based RAG-as-a-Service platform serving 20+ business units, abstracting data ingestion and search infrastructure, cutting time to build first MVP from weeks to under a day.
• Built production-ready multi-agent RAG using LangGraph with structured outputs, guardrails and retry mechanisms, and deployed an MCP server for LLMs to leverage RAG platform as a tool for agents.
• Designed evaluation framework for RAG platform with golden-set benchmarks, LLM-as-a-judge scoring, and reproducibility across the enterprise platform.
• Established robust end-to-end API tests in CI/CD pipelines and observability practices for request traces, alerts and latency monitoring.
• Shaped the product roadmap for the RAG-as-a-Service platform, scope and prioritize features in collaboration with multiple business stakeholders and platform architects.
• Implemented data ingestion pipelines for RAG to leverage metadata for hybrid search and filtering, to integrate critical internal data sources into Azure AI Search to feed LLM applications.
• Identified and optimized redundant search service infrastructure through metadata-based index consolidation cutting costs by 75% (saving €250k+ annually) with no regression in retrieval quality.

Machine Learning Engineer Sep 2022 – Feb 2025

• Automated versioning and multi-environment deployments with GitLab CI, FluxCD, and Helm, eliminating manual release overhead.
• Provisioned scalable AWS EKS clusters and storage with AWS CDK, delivering identical, production-ready environments.
• Orchestrated automated ML training and evaluation pipelines with Prefect and MLflow, improving experiment tracking and collaboration.
• Standardized Docker images and DevSpace templates, reducing developer onboarding from days to hours.
• Designed and ran technical training workshops to upskill the team on newly adopted infrastructure and workflows.

Senior AI Researcher Jan 2022 – Aug 2022

• Developed a novel time-series domain adaptation algorithm to scale predictive models across machines, securing US Patent No. US12032929B2.
• Designed automated root-cause and prescriptive-analysis AI features with product teams and domain experts.
• Supervised three AI researchers and an intern building scalable data processing algorithms for plant-wide AI rollouts across two manufacturing facilities.
• US Patent · US12032929B2 · Granted Jul. 2024 — System and Method for Cross Domain Generalization for Industrial Artificial Intelligence Applications
• EP Patent · EP4254087B1 · Granted Mar. 2026 — System and Method for Recommending a Recipe in a Manufacturing Process
• EP Patent · EP4300229A2 · Application Pending — System and Method for Determining Defect Regions of Products in a Manufacturing Process

AI Researcher Sep 2020 – Dec 2021

• Deployed a predictive model prescribing optimal casting parameters for a wheel manufacturer, reducing scrap by 50% and saving €144k annually.
• Developed an automated time-series framework standardizing data integration and feature engineering, cutting deployment timelines from 50 to 30 days.
• Built proofs-of-concept for 5+ predictive maintenance and quality-optimization use cases, driving AI adoption for manufacturing clients.

Research Assistant — Machine Learning Feb 2019 – Aug 2020

• Developed deep learning algorithms and feature pipelines for multivariate sensor data, optimizing computer vision and predictive maintenance systems.
• Developed a novel regularization technique and multi-headed neural network, reducing catastrophic forgetting from 6.5% to 1.0% on sequential classification tasks.