Iβm an AI Research Engineer and former Researcher at IISc Bangalore, working at the intersection of multimodal AI, misinformation detection, and agentic AI systems. My experience spans both academic research (WACV 2025, IIT Jodhpur) and industry-driven innovation (Johnson & Johnson, DRDO, Google, KissanAI), where I focus on building deployable, high-impact AI solutions.
- Multimodal Fake News Detection (WACV 2025)
- Deepfake Detection on Fully & Partially Generated Images
- Vision-Language Integration for ASR (Google Vaani Project)
- Dataset curation for Indian misinformation landscape
- Optimizing inference pipelines using vLLM and LangGraph
- Languages: Python, SQL
- Frameworks: PyTorch, Hugging Face, LangChain, LangGraph
- ML & DL: CV, NLP, Multimodal Learning, Fake News Detection, Deepfakes
- LLMs: RAG, PEFT, SFT, Agentic Workflows
- Deployment: FastAPI, Streamlit, Docker, CI/CD
- Other Tools: Git, Docling, SHAP, LIME
π Built Innovation Graph β a GraphRAG-powered platform for collaborative ideation (Johnson & Johnson, AAV Labs) π Published at WACV 2025 on multimodal misinformation detection (91.4% accuracy, SOTA) π§ Developed agentic pipelines for inference & reasoning using LangGraph (10Γ speedup) π Designed deepfake detection system (98% full, 64% partial generations) π€ Collaborated with Google Vaani to improve ASR WER from 25 β 21 πΎ Fine-tuned Qwen-7B for agriculture-based VQA (KissanAI, Hugging Face release)
- FRAUD-Net: Multimodal Fake News Detection using Uncertainty & Domain Adaptation (WACV 2025)
- Gravitational Wave Classification using GMM and GLM (2023)
- Portfolio Website
- π§ [email protected]
Thanks for stopping by!