DVC Experts in Germany

• Developed end-to-end reproducible ML pipelines (PyTorch) with data versioning (DVC), experiment tracking (MLflow), automated testing (PyTest), and CI/CD across all training workflows.
• Scaled Vision Transformer and CNN training across NVIDIA A100 GPU clusters (CUDA, DDP, SLURM); applied hyperparameter optimization (W&B Sweeps) to reduce training overhead and identify optimal configurations.
• Developed a real-time 3D human motion generation system (ViT, VQ-VAE, SMPL-X/PIXIE) for personality-conditioned avatar synthesis; achieved state-of-the-art FID = 6.15 and P-FID = 10.31 on the UDIVA benchmark.
• Validated model expressiveness through structured user studies, achieving 86% accuracy in distinguishing extroverted vs. introverted avatar behaviors.
• Optimized inference pipelines by deploying PyTorch models via TensorRT and ONNX Runtime into native C++ code; benchmarked performance.

• Building and scaling cloud infrastructure on GCP to support a SaaS platform for industrial clients
• Designing and implementing a data-driven DevOps pipeline for streamlined deployment and CI/CD workflows
• Collaborating with Data Science team on MLOps workflow to automate integrated retraining

• Design of an LLM-based agent system to control the ERP software
• Development of agent workflows with LangGraph and PydanticAI
• Planning interfaces between business logic and language models
• Planning agent orchestration
• Prototype development and demonstration

Tools: Python, Pydantic, React, LangChain, LangGraph, Linux

From Crude to Crafted: Refining Participatory Design Data into Stakeholder-Ready Outcomes

• Architected a production Document AI platform using Retrieval Augmented Generation (RAG) over 1,500+ participatory design artefacts to answer historical project queries with grounded responses.
• Designed LLM evaluation combining RAGAS, custom evaluation metrics and human-in-the-loop (HITL) validation workflows to evaluate factual grounding, response quality, and prompt performance.
• Built a React, TypeScript, and D3.js frontend for interactive exploration of AI-generated insights.
• Implemented input layer LLM safety controls and Guardrails, including PII redaction and foul language filtering.

• Developed and optimized LSTM-RNN and Decoder Transformer models to predict vehicle trajectory during target loss events in Adaptive Cruise Control systems, achieving 20% improved predictive accuracy over baseline models.
• Engineered novel data preprocessing pipeline from real road campaign data, processing multi-sensor time series data, generating 300+ training snippets.
• Implemented Bayesian hyperparameter optimization and applied physical constraints to prevent model run-away behavior, resulting in 30% smoother acceleration profiles.
• Extended existing patented technology for AI-assisted ACC function improvements, building upon foundational work to enhance network performance.
• Tools: Python, TensorFlow, Keras, Optuna, CarMaker

• Led a team of junior Research Engineers, providing mentorship, technical guidance, and career development support to foster their growth in deep learning and machine learning engineering.

• Developed LLM-based automation utilities including structured reasoning pipelines, LLM-as-a-Judge evaluation tools, and multi-model comparison frameworks.
• Built RAG pipelines for internal research workflows using LangChain, ChromaDB, and FastAPI, enabling semantic retrieval and multi-step reasoning.
• Integrated LLM microservices into existing ML systems using Docker, FastAPI, and GitLab CI/CD with reproducible deployment workflows.
• Designed inference APIs combining vision models and LLM reasoning for multimodal analytics and decision-making.
• Optimized embedding-based retrieval using vector store pruning, improved chunking logic, and dynamic retriever selection.
• Performed prompt engineering and system instruction tuning for consistency, robustness, and reasoning quality.
• Built benchmarking suites to evaluate LLM latency, reasoning quality, retrieval accuracy, and robustness under different prompt templates.

• Built data pipelines with DVC for version control and efficient data management
• Trained and optimized AI models
• Improved CVAT with custom annotation formats, AI model integration, and streamlined annotation workflows
• Trained, debugged, evaluated, and deployed DCNN models in production
• Developed MaDCAT, an AI-powered CVAT extension for simultaneous data capture and annotation

• Designed and implemented end-to-end automated workflows for extracting structured data from semi-structured PDF documents including invoices and medical reports
• Leveraged Optical Character Recognition (OCR) technology and large language models to parse documents and generate validated JSON schemas
• Engineered prompt optimization strategies and rule-based classification hierarchies to enhance parsing accuracy across diverse document layouts
• Established quality assurance framework using evaluation metrics to validate output against ground truth datasets with 96% accuracy