I build production-grade AI systems and scalable backends — and research how intelligent agents can learn from minimal signals.
Currently:
- 🏗 Architecting AI infrastructure and agent systems in production
- 🧠 Researching sample-efficient offline reinforcement learning
- 🌍 Designing world-model pipelines for robot learning
- ⚙️ Building systems that move from research → deployment
Founding Engineer, Backend & AI — Confyde.ai
- Architected and deployed scalable AI backend using FastAPI + Supabase
- Designed distributed task orchestration with Celery + GCP Pub/Sub
- Built LangChain / LangGraph-based research agents for clinical and market intelligence
- Implemented CI/CD pipelines on GCP with GitHub Actions
- Delivered production AI systems processing thousands of weekly computations
Core Focus Areas
- AI agents & orchestration
- API and systems design
- Distributed workers & async processing
- Data pipelines & biostatistical modeling
- Cloud-native deployments (GCP, AWS)
NYUAD Deep Learning Lab — Deep Learning Researcher
- Learning state-only representations for robot control
- Implementing offline RL algorithms (ReBRAC, TD3+BC, IQL)
- Designing latent world models for data-efficient reinforcement learning
- Exploring sim-to-real transfer and policy evaluation
Research themes:
- Sample-efficient offline RL
- JEPA-style architectures for RL
- Representation learning for control
- Planning from state-only signals
Languages: Python, C/C++, TypeScript
ML: PyTorch, Representation Learning, World Models, LLMs, RAG
Backend: FastAPI, Flask, API Design, Celery, CI/CD
Infra: PostgreSQL, MongoDB, Docker, Linux, Git
Cloud: GCP, AWS
