Loop
AI-powered study planner that builds personalized learning paths based on a student's goals, pace, and deadlines, using LLMs to generate adaptive schedules and quizzes.
- React
- Firebase
- Python
- OpenAI
- FastAPI
AI-powered study planner that builds personalized learning paths based on a student's goals, pace, and deadlines, using LLMs to generate adaptive schedules and quizzes.
Food safety and nutrition inspector that uses Gemini to read labels and ingredients, flag allergens and additives, and surface personalized nutrition insights.
Silver medal Kaggle solution demonstrating robust prompt-injection and multi-step tool attacks. Deployed a stateful 'Confused Deputy' pipeline that bypassed hidden provenance/taint-tracking guardrails while optimizing container throughput.
How the desktop app's Local/SSH/Remote sessions quietly replaced a VPS, tmux, and Tailscale setup — and the few cases where a real terminal still earns its keep.
How LLM-based agents observe, reason, and act: model, tools, and orchestration layer, plus ReAct/CoT/ToT with a LangChain example.
Text, image, and multimodal embeddings, vector databases like Pinecone and Chroma, and how ANN search (HNSW, FAISS) makes retrieval fast.
Temperature, top-K, and top-P, plus the prompting techniques that actually move the needle: zero-shot, few-shot, Chain of Thought, ReAct.
From the 2017 Transformer to GPT-4 and Gemini: pretraining, fine-tuning with SFT/RLHF/LoRA, and inference optimization.
Senior Data Scientist with 7+ years of experience, currently leading GenAI initiatives across the Glovo and Delivery Hero (DH) ecosystem. I focus on productionizing multi-agent architectures and automated taxonomies that serve as the foundational data layer for millions of users globally.