Vicky Gupta🇩🇪

AI Researcher

Erlangen, Bavaria, Germany6 mos experience

AI EnabledAI ML Practitioner

Key Highlights

Developed innovative LLM Agent for medical imaging.
Specialized in GenAI and Retrieval-Augmented Generation.
Eager to drive transformative AI innovations.

Stackforce AI infers this person is a Healthcare-focused AI Developer with expertise in Generative AI and Large Language Models.

Contact

Skills

Core Skills

Agentic AiRetrieval-augmented Generation (rag)Large Language Models (llm)

Other Skills

Python (Programming Language)Large Language Model Operations (LLMOps)Microsoft AzureAzure AI FoundryLeadershipTeam LeadershipManagementLangChainLangGraphAI AgentsVector DatabasesHigh Performance Computing (HPC)Prompt EngineeringNatural Language Processing (NLP)Manual Testing

About

From the moment I first explored the world of AI in 2018, I knew I had found my passion. I’m Vicky Gupta—a soon-to-be M.Sc. graduate in Artificial Intelligence from FAU Erlangen-Nürnberg, specializing in GenAI, Large Language Models (LLMs), and Retrieval-Augmented Generation (RAG). My journey has been fueled by a drive to simplify complex systems and make them accessible to everyone. Currently, as a Research Student in GenAI at SOHARD GmbH, I am developing an innovative LLM Agent that converts natural language directly into SPARQL queries, enabling our PACS System (SEDI PACS) to use SPARQL directly for querying graph databases in medical imaging. By leveraging prompt engineering, fine-tuning, and high-performance computing (HPC) on NVIDIA A100 GPUs, I am continuously working on implementing better fine-tuning methods with a validation framework to improve accuracy to a level where this system can be deployed in production. I’m eager to leverage my skills and achievements in a full-time GenAI role, contributing to transformative AI innovations. Let’s connect and explore how I can drive the next wave of AI excellence in your organization! #GenAI #LLM #RAG #ArtificialIntelligence #MachineLearning #Innovation #AIJobsGermany

Experience

6 mos

Total Experience

6 mos

Average Tenure

Current Experience

Fraunhofer iisb

Agentic AI Developer

Oct 2025 – Present · 7 mos · Erlangen · On-site

Contributing to engineering and on-premise deployment of the SimAgent, an internal agentic AI research platform
Implemented multi-agent workflows using LangGraph and LangChain, defining agent coordination and research orchestration
Perform full-stack debugging to identify and resolve bugs in both frontend and backend components, enhancing system stability
Developed tool-calling capabilities enabling research agents to dynamically invoke tools, significantly improving research depth
Deployed locally hosted GPT-OSS 120B model to meet data privacy and security requirements without relying on external APIs
Handled production failures and server failures throughout the development process to minimize the risk of common failures

Python (Programming Language)Agentic AI

Neureya labs

AI Engineer

Aug 2025 – Oct 2025 · 2 mos · Germany · Remote

Contributed to the development of VirtuallyU, a next-generation AI platform transforming real individuals into interactive, lifelike AI avatars.
Designed and implemented a Retrieval-Augmented Generation (RAG) agent on Azure AI Foundry, enhancing LLM performance through contextual knowledge integration.
Implemented LLM guardrails to ensure safe, reliable, and policy-compliant outputs across the AI platform.
Improved response quality, contextual grounding, and overall system robustness in a production environment.
Actively participated in cross-functional strategy and product meetings, proposing AI development methodologies to enhance system performance and development efficiency.
[Note]: Company operations ceased due to funding constraints and high infrastructure costs associated with large-scale image generation and model inference.

Python (Programming Language)Retrieval-Augmented Generation (RAG)

Sohard

2 roles

GenAI Developer | NL-to-SPARQL (RAG Agent)

Promoted

Jan 2025 – Jul 2025 · 6 mos · On-site

Master's Thesis : LLM-Centric Framework for Ontology-Driven SPARQL Query Generation in RAG for DICOM Databases
Developed an LLM-powered RAG system to automate SPARQL query generation from natural language, for DICOM Database.
Fine-tuned LLMs on hand-curated datasets and built prompting strategies, optimizing semanticity and query generation accuracy.
Integrated proprietary SOHARD GmbH’s Ontology into the RAG framework, ensuring domain-specific validation and accuracy.
Optimized HPC workflows on clusters using NVIDIA A100 GPUs, enhancing efficiency for large model training and inference.
Boosted SPARQL query accuracy to 70% on Llama 70B via ontology-RAG integration on SOHARD’s PACS (SeDI).
Reduced evaluation time by 90% by developing an LLM-powered Evaluation Agent significantly improving efficiency.
Containerized workflows with Docker, enabling reproducible deployments across HPC and dev environments

Large Language Models (LLM)Large Language Model Operations (LLMOps)

GenAI Developer | NL-to-JSON (LLM Agent)

Jul 2024 – Dec 2024 · 5 mos · On-site

Project 1 : Natural Query to JSON Project (LLM, GenAI, RAG Agent)
Developed an LLM Agent that converts natural language queries into JSON format for querying a graph database.
Integrated AI-powered chat functionality into PACS software, allowing users to retrieve medical imaging data using conversational queries instead of manually inputting variables, improving efficiency and planning.
Implemented Prompt Engineering, Retrieval-Augmented Generation (RAG), Vector Databases, Function Calling, Fine-tuning, and Memory Integration to enhance LLM performance.
Designed and hand-curated fine-tuning datasets to improve model accuracy.
Conducted experiments with multiple LLMs, including LLaMA 3.1, Mixtral, DeepSeek, and Olmo.
Successfully delivered a working solution with significant accuracy, improving automation in medical data retrieval.

Retrieval-Augmented Generation (RAG)Large Language Models (LLM)

Pattern recognition lab

Master Thesis Project

Jan 2025 – Jul 2025 · 6 mos · Erlangen · Hybrid

LLM-Centric Framework for Ontology-Driven SPARQL Query Generation in RAG for DICOM Databases

Internity foundation

ML Intern

May 2020 – Jul 2020 · 2 mos · India

During this internship, I learned lots of practical life algorithms of ML and there mathematics intuition and the python implementation. At last, we made a project on which basis we are evaluated for letter of completion.