Software Engineer bridging AI systems and embedded development. Building intelligent, high-performance systems that translate complex technical challenges into measurable outcomes.
A Software Engineer with deep expertise in AI/ML systems and embedded development, with hands-on experience across the full Agile/SCRUM delivery cycle.
A natural problem-solver bridging research and real-world engineering — focused on building intelligent, high-performance systems that streamline workflows.
Passionate about automotive technology, from surface crack detection to DDoS protection in vehicle networks and AI-driven predictive maintenance.
Currently pushing boundaries at Nokia, Bangalore, automating network performance testing and integrating CI/CD pipelines for faster releases.
Ensemble learning model combining CNN, Random Forest, KNN and XGBoost to improve crack detection accuracy. Implemented data augmentation to reduce false positives on diverse industrial component surfaces.
AI-based predictive maintenance using Random Forest and LSTM models to detect early equipment failure. Continuous monitoring via AWS IoT and REST APIs.
AI-based system for detecting DDoS attacks in software-defined automotive networks, ensuring secure and reliable vehicle communications.
Built using LangChain and LLMs, enabling natural language interaction with German-language PDF documents for efficient information retrieval and summarization.
Fine-tuned Llama-3 using LoRA adapters on domain-specific instruction datasets. Evaluated performance using BLEU, ROUGE, and human preference evaluation. Reduced GPU memory consumption using QLoRA.
Built a large-scale entity-resolution platform using semantic embeddings, approximate nearest-neighbor search, and graph-based similarity to identify duplicate addresses across millions of records. Designed scalable retrieval, matching, and clustering pipelines to construct canonical address hierarchies.
Developed a graph-based platform for delivery-location prediction, ETA estimation, and delivery-success prediction using Graph Neural Networks and OpenStreetMap data. Integrated geocoding, road-network analysis, route optimization, graph embeddings, predictive modeling, and scalability benchmarking.
Developed an end-to-end system for noisy, incomplete, and multilingual address inputs using normalization, entity extraction, semantic matching, validation, and confidence scoring. Designed modular NLP pipelines using BERT, DeBERTa, IndicBERT, and embeddings.
Open to opportunities, collaborations, and conversations. Let's build something extraordinary.
Have an idea? I'd Love To Hear It.