Roshan Rayala Bhaskar
|
Digital Engineering specialist passionate about merging technology with real-world challenges through Generative AI and Machine Learning.
About Me
I’m an AI Engineer with a background in product design, manufacturing and digitalization, focused on building practical, production-ready AI systems for real industrial challenges.
My work spans Computer Vision, 3D applications, Vision-Language Models, RAG systems, and ML pipelines used in engineering and manufacturing domains. I enjoy working on the full lifecycle of an AI solution — from understanding technical requirements to designing models, building scalable pipelines, and deploying them in real environments.
I’ve developed solutions such as Automated analysis of Technical Documents using hybrid LLM Frameworks, multimodal retrieval systems for technical document QA, prediction models for configuring complex fuel-cell test stations, OCR workflows, and end-to-end MLOps systems with MLflow, Docker, and CI/CD.
Across projects, my goal is simple: use AI to reduce manual effort, improve decision-making, and unlock new levels of efficiency in engineering processes.
Languages
Technical Skills
Programming
ML & DL
LLMs & Frameworks
MLOps & Software
Data Engineering
Cloud
Soft Skills
Experience
Master Thesis - GenAI & Process Automation
ATA Anlagentechnik Aschersleben GmbH | OVGU Institute- Built a hybrid AI pipeline with vision-language models and synthetic data to automate symbol and attribute extraction from CAD drawings for BOM and welding workflows.
- Optimized multimodal models for lightweight hardware using quantization techniques, with research accepted for publication in Operations Management Research.
Horiba FuelCon GmbH
Magdeburg, Germany
Data Science
- Developed ML pipelines for predicting system-tags, built OCR tools for extracting data from CAD documents.
- Created PoC text-to-SQL systems for efficient querying of engineering databases.
Data Analytics
- Developed a full-stack analytics platform integrating SQL data and CI/CD automation.
- Improved real-time test station analysis and streamlined internal workflows.
- Supported ML pipelines and contributed to an image-to-3D proof-of-concept for early automotive design workflows.
- Built CAD/PLM software tools to streamline engineering tasks and improve design productivity.
- Spearheaded product design and development for manufacturing sectors.
- Optimized production workflows and improved design efficiency.
Education
Master of Science - Digital Engineering
Otto-von-Guericke University, Magdeburg, Germany
- Focus: Generative Models, Computer Vision, Deep Learning
- Grade: 2,1
Postgraduate Diploma in AI & ML
University of Hyderabad, India | Online-Plattform
Bachelor of Mechanical Engineering
Visveswaraya Technological University, Hassan, India
- Focus: Product Design and Manufacturing, CAD, C Programming, Engineering Management
- Grade: 1,9
Featured Projects
Video Compliance Evaluation in Azure
A robust compliance pipeline that audits YouTube videos against the YouTube Child Safety Policy. Built with LangGraph orchestration, Azure Video Indexer for transcripts and OCR, Azure AI Search for policy-rule retrieval via RAG, and GPT-4o through FastAPI to generate structured safety reports visualized in Streamlit, with tracing via Azure Application Insights and LangSmith.
End-to-End MLOps Pipeline with MLflow & CI/CD on AWS
Designed and implemented a modular, end-to-end MLOps pipeline integrating data stages (ingestion, validation, transformation) with ElasticNet model training and leveraging MLflow for experimentation tracking; deployed Flask app to AWS EC2 using Docker and GitHub Actions.
Autodesk Alias API MCP Server
Built an MCP server that makes Autodesk Alias API documentation searchable inside MCP clients. The system indexes 231 scraped documentation pages with BM25 ranking and exposes validated tools to search docs, fetch full pages by title, list docs with pagination, and retrieve code examples for faster plugin and API development workflows.
Multimodal RAG Application for Custom Document QA
Built a scalable multimodal RAG system (LangChain + LLMs + FastAPI) enabling end-to-end processing of text, image, and tables in documents for practical, industrializable QA workflows.
LLM Framework for Automated Quality Check in Requirements Engineering
Created an OpenAI LLM-powered assistant for analyzing and revising natural language requirements according to 42-INCOSE rules, with automatic classification and interactive visualization in Streamlit.
OCR-Enhanced CAD Data Extraction and Processing
Developed an interactive app for dimension detection and extraction in CAD drawings using OpenCV and EasyOCR, achieving 96% accuracy with Human-in-the-Loop following Industry 5.0 principles.
Unity VR Training Application for Neurosurgeons
Gamification-based VR training app for neurosurgeons to analyze DSA images. Built with Unity and SteamVR, providing an immersive 3D environment for interacting with cerebral vasculature models and enhancing visuospatial skills.
Deep Learning and Generative Models Collection
Comprehensive collection of DL & Generative Models implementations including MLPs, CNNs, RNNs, Transformers, VAEs, GANs, Autoregressive Models, NICE, Score-Based Models, and Denoising Diffusion using TensorFlow & PyTorch.
Publications
Bridging Manual & Automated Workflows: CAD Drawing Data Extraction in Industry 5.0
Journal Article
Extra Curricular Activities
Creative Head
OVGU Academic ClubAs the Creative Head at the OVGU Academic Club, I was responsible for orchestrating a variety of technical and non-technical events, fostering collaboration among peers and faculty, and contributing to a culture of creativity, intellectual curiosity, and academic excellence.
We organized impactful events including Recruiters Event, Master Your Thesis, PhD – Why, What, and How, as well as technical workshops like Introduction to MLOps and Dock Your Docker series. These events were designed to help students gain valuable insights for their careers, academic pursuits, and professional networking. Working as a cohesive team, we took on individual responsibilities that allowed us to develop strong leadership and collaboration skills. In my role, I focused on creatively organizing these events, driving outreach and advertising efforts to maximize engagement among students, faculty, and industry partners.