München, Deutschland · LinkedIn · prateekgaur@gmx.de · Portfolio

I am an Applied ML Engineer with a background that grew from practice: Mechanical Engineering (B.Tech.) → Battery Systems & Energy Engineering (M.Sc. TU Berlin) → Machine Learning & AI Engineering

This combination is rare. I don't just understand Python, PyTorch, and LLM pipelines — I understand what the data physically means. Battery degradation models, thermal simulations, time-series data from industrial systems: this is the domain knowledge my ML solutions are built on.

I build multi-agent systems with human-in-the-loop review, production-grade computer vision pipelines, and deployed RAG/predictive-maintenance applications — bringing the same rigor from battery engineering to agentic AI architecture. Several projects below are live and clickable, not just described.

Predictive maintenance dashboard for solar inverter fleets — fleet-level anomaly detection, alert breakdowns, and unit-level drill-down, all in a deployed, interactive interface. 🔗 Live demo Python Streamlit Time-series Predictive Maintenance

RAG-based assistant for tender and grid compliance document review — semantic search and Q&A over regulatory documents, deployed on Streamlit Cloud. 🔗 Live demo) Python LangChain RAG Compliance

End-to-end platform merging extraction, anomaly detection, classification, and decision agents with a confidence-based human review workflow. Processes records through automated quality gates, escalating low-confidence cases for review and feeding corrections back into an RLHF dataset. Full audit trail and live dashboards. Python Streamlit n8n PyTorch FAISS Ollama scikit-learn HITL

Battery cell defect detection system built on a custom CNN (98.8% test accuracy, 0.988 macro F1) orchestrated by four cooperating agents — vision, memory, decision, and feedback — communicating over a typed Agent-to-Agent (A2A) protocol. Includes cross-session memory and live drift detection. Python PyTorch OpenCV A2A Protocol Multi-Agent Computer Vision

German-language automated accounting and royalty-reconciliation tool for franchise operations — anomaly detection across franchise fleets and automated reconciliation reporting. Python German-language Automation Anomaly Detection

German-language toolkit for structuring and automating client onboarding workflows for non-technical business teams. Python German-language Workflow Automation

LSTM/GRU/Transformer ensemble for SOH/SOC/RUL prediction with FastAPI serving, Streamlit dashboard, Evidently AI drift monitoring, and Kubernetes HPA auto-scaling (1→5 pods); CI/CD via GitHub Actions to Azure and AWS. Python PyTorch FastAPI Kubernetes Docker Evidently AI CI/CD

Production-ready Retrieval-Augmented Generation pipeline over internal simulation guidelines. Integrates local LLMs for data privacy. Semantic search over complex engineering specifications. LangChain LlamaIndex FAISS RAG local-LLM

LSTM and RNN models for State-of-Health estimation and degradation forecasting on real industrial battery datasets. Benchmarked against Random Forest and SVM baselines (~18% lower MAE, MAE=0.018, R²=0.94). Python PyTorch LSTM time-series battery

🎓 M.Sc. Energy Engineering — Technische Universität Berlin (2021) Thesis: Custom Battery Cell Balancing Circuit Design Under Thermal Gradient

🎓 B.Tech. Mechanical Engineering — Rajasthan Technical University (2016), First Division

✅ Machine Learning Specialization — DeepLearning.AI / Stanford Online ✅ Battery Management Systems — University of Colorado Boulder (with honours) ✅ Python for Everybody — University of Michigan ✅ CATIA V5 Certified Associate — Dassault Systèmes ✅ Siemens NX Certified Designer

ML Engineer · AI Engineer · Data Scientist · Battery Systems Engineer Open to roles in Munich and remote across Germany

Languages: English (C2) · German (B2, actively improving) · Hindi (native)