1
1
United States
5+ years of experience
I have hands-on experience building end-to-end AI systems, primarily focused on machine learning and retrieval-augmented generation (RAG). I design pipelines for document ingestion, preprocessing, embedding generation, and semantic search, integrating them with LLMs for reasoning and automation. My systems typically use modular architectures (e.g., FastAPI backends, Next.js frontends) and support workflows like document analysis and interactive review. I’ve implemented classical ML models (logistic regression, random forests) and ensemble methods, with a strong focus on evaluation using metrics like PR-AUC and out-of-sample testing. I also explore agent-based systems (LangChain, DSPy) to automate workflows and enable dynamic tool use. Additionally, I’ve worked on quantum–AI hybrid models using Qiskit, including quantum kernels (QSVC) and noise-aware benchmarking. My approach emphasizes scalable, production-ready systems with clean APIs, containerization, and extensible design.
.png&w=828&q=75)
Agent-to-Agent Compute Marketplace is a trustless compute market for the agentic economy. Consumer agents request work such as image classification, model inference, embeddings, or data processing from provider agents. Each request is metered, priced, checked against a hard per-unit ceiling of $0.01, and settled in USDC through an Arc smart contract. The project includes a Vyper marketplace contract, a FastAPI backend with metering, pricing, settlement, provider registry, and reputation engines, plus a Next.js UI for stakeholders. The UI lets users register providers, submit compute jobs, run batch jobs, inspect backend health, and view provider reputation. The demo proves why nanopayments matter: traditional gas costs make per-action compute payments uneconomical, but Arc + USDC nanopayments make sub-cent autonomous agent settlement viable. The local demo can generate 50+ agent-to-agent transactions in simulated mode, and the backend is structured to switch to live Arc settlement when funded testnet credentials are configured.
26 Apr 2026