I build Agentic AI
and LLM systems that ship.

• → Agentic AI: Designing multi-agent systems (LangGraph) that automate complex analytical and reporting workflows — orchestrating agents for information gathering, analysis, insight generation, and evidence-backed business reporting
• → LLM & SLM Engineering: Fine-tuning domain-specific language models via SFT, LoRA/PEFT, QLoRA and preference optimisation (DPO, GRPO); covering training-data preparation, custom reward pipelines, evaluation, and inference optimisation on multi-GPU (A100/H100) infrastructure
• → Building production-ready GenAI solutions at the intersection of Agentic AI, applied ML, and financial services

• → Transformer-based drug shortage prediction — detected disruptions 1–2 months earlier
• → Bayesian MMM + budget optimizer — estimated +20–25% revenue uplift
• → XGBoost pharmacy forecasting with E2E AWS deployment + OpenAI-powered explanations

• → Predictive lead scoring on SageMaker — improved conversion targeting for pharma sales
• → Gate-checked bags forecasting for US airline — 10–15% fewer under-predictions
• → NLI pipeline (PubMedBERT on Databricks) for medical insight mining

Gaze estimation iOS app and driver safety demo using convolutional models.

Embedded systems and product engineering for automotive and broadcast technology domains.