Student Assistant
JAN 2025 – PRESENT
Fraunhofer IEM, Institute for Mechatronic Systems Design, Paderborn
- Led 5- and 3-person teams across two airport-automation projects. Delivered the first on schedule, which earned the follow-on scope.
- Designed a 10-microservice backend on FastAPI, Kafka, and Cassandra with Redis for hot state, isolating failures and scaling services independently.
- Integrated Blickfeld (LiDAR), Gemini (stereo camera), and Lupus sensors on-site. Wrote reusable calibration procedures that cut setup time on future installs.
- Fine-tuned a computer-vision model on airplane components (nose, wings, engines) with segmentation-model fusion for real-time localisation, halving error from 10 cm to 5 cm and eliminating misalignments.
- Engineered sensor-driven luggage dimensioning that feeds the autonomous robot-loading stage.
- Provisioned on-site servers and remote access, enabling 24/7 monitoring of deployed systems.
Student Research Assistant
NOV 2023 – OCT 2024
DICE Research Group, Paderborn
- Co-developed the Materials AI Agent, a semantic search interface for Springer Nature, with work published at ESWC 2024 (Springer).
- Pretrained a BERT MLM on material-science vocabulary, powering domain-aware language understanding inside the chatbot.
- Built a LangChain tool that indexes input text into vector storage, the retrieval backbone for downstream RAG queries.
- Filtered RAG retrieval by indexed URIs, improving retrieval precision by 78%.
- Fine-tuned and benchmarked Mistral, Zephyr, and Llama2 variants on material-science NER, reaching a 90% F1 score.
- Integrated an agent system into the
ai-search backend route, powering typeahead prediction of the next compound or material-science term as the user types.
Data Science & ML Scholar Intern
JAN 2022 – FEB 2022
Tathastu Technologies Pvt. Ltd, Noida, India
- Exploratory data analysis across multiple datasets and feature engineering that lifted model accuracy.
- Applied ML techniques to five real-world problem statements end to end.
