Tanuj Kumawat

I’m a B.Tech Chemical Engineering student at IIT Roorkee, focused on building applied AI and data solutions that perform in the real world. I enjoy translating messy signals into models, pipelines, and decisions teams can trust.

As an AI Research Intern in the ECE Department, I analyzed multimodal physiological data (EEG, ECG, EMG, GSR) for driving behavior analysis and proposed a multi-encoder architecture with LSTM-based attention per modality. Using multi-head attention for modality fusion, I trained on 30 subjects and achieved 70% accuracy on 5 driving patterns.

In my projects, I built a SQL & Python pricing intelligence engine using OLS regressions to prevent a 54% SKU underpricing error, and I developed an agentic LinkedIn post generator with Llama 3.3 and FastAPI on Vercel—cutting backend overhead by 66% with quality and safety guardrails. I’ve also worked on graph-based ETA modeling (reducing prediction errors by 22.5% and reaching 75.7% SLA accuracy), and marketing attribution/portfolio optimization to eliminate bot-wasted spend and improve conversion outcomes.