talha nasir

I've built production AI and SaaS systems for US and European companies, most recently at Bespoke Labs, where I own full-stack delivery for Curator, a post-training platform for frontier AI lab clients. I ship React, Next.js, and FastAPI features alongside PyTorch and TensorFlow pipelines, automated RLHF scoring, and ETL workflows. I also designed RAG systems using Pinecone, Weaviate, and FAISS to improve retrieval relevance, reduce hallucinations, and optimize inference costs. At Bespoke Labs, I built a secure multi-tenant knowledge base chatbot with Python microservices, LangGraph orchestration, GCP, Docker, Kubernetes, OAuth, and API protection. I partner with research leads on reusable optimization tooling and monitor performance, reliability, and infrastructure costs with Prometheus and Grafana. Previously at Devsinc, I led backend delivery for three SaaS products, maintaining 99.8% uptime while improving throughput by 21% and reducing QA-reported issues by 18% through AI-assisted automation. Across Rails, Python, Node.js, React, AWS, and CI/CD systems, I work end-to-end and asynchronously across time zones.