Victor – Python, LLM, AI agent development

A decentralized AI infrastructure and confidential compute network for verifiable AI execution, privacy-preserving machine learning, and GPU-powered proof systems. It combines high-performance GPU compute, zero-knowledge proofs, and trusted execution environments to give developers and enterprises cryptographic verification, confidential execution, and transparent proof-of-compute for AI inference and agent workloads.

• Architected and built core infrastructure for the protocol using Rust, CUDA, Cairo, TypeScript, and Next.js;
• Designed and implemented GPU-accelerated ZK proving systems for machine learning inference using STWO (Circle STARKs) and GKR sumcheck protocols;
• Developed high-performance CUDA kernels and proving pipelines for large-scale neural network verification workloads;
• Built and deployed the first on-chain verification workflow for a 14B-parameter neural network (Qwen2.5-14B) on Starknet Sepolia;
• Created distributed AI compute and proof orchestration infrastructure for GPU workers and confidential compute environments;
• Developed smart contract verification systems and Cairo-based proof verifiers integrated with Starknet;
• Built production-ready developer tooling, including SDKs, APIs, marketplace infrastructure, and automation systems across Rust, TypeScript, Python, and Next.js ecosystems;
• Led architecture decisions around verifiable inference, AI security, privacy-preserving computation, and proof-gated compute execution;
• Published open-source tooling across crates.io, PyPI, and npm, including the Obelyzk SDK and supporting developer libraries;
• Contributed to protocol design, validator systems, compute marketplace architecture, and decentralized infrastructure strategy;
• Represented the project publicly through technical demos, developer relations, and conference speaking engagements, including zkSummit 14 in Rome.

An AI data intelligence platform that helps teams turn operational data into real-time insights and automated workflows. It combines GPU-backed agents, LLM tool use, vision models, object detection, anomaly detection, and an NL-to-SQL interface over operational data lakes.

• Architected and shipped the full production platform;
• Designed the GPU-backed agent system with LLM tool use, transformer-based vision models, YOLO, DETR, real-time anomaly detection, and an NL-to-SQL interface over operational data lakes.

An open-source ZKML and verifiable AI infrastructure protocol for proving large-scale machine learning inference with GPU-accelerated zero-knowledge systems. It combines CUDA acceleration, STARK-based proving, Cairo smart contracts, and distributed proof infrastructure to make AI execution cryptographically verifiable, auditable, and trust-minimized.

• Authored the open-source Rust ZK proving engine for ML inference;
• Built one of the first on-chain verification systems for a 14B-parameter neural network (Qwen2.5-14B) on Starknet Sepolia;
• Designed and implemented the first GKR sumcheck prover on StarkWare’s STWO framework;
• Developed high-performance CUDA kernels and GPU proving infrastructure for large-scale neural network verification workloads;
• Wrote and maintained 20,000+ lines of custom CUDA and proving infrastructure code, backed by 950+ automated tests;
• Built Cairo-based smart contract verification systems integrated with Starknet for proof validation and on-chain execution verification;
• Architected distributed proof orchestration systems for scalable AI proving and verification pipelines;
• Developed SDKs and developer tooling published across crates.io, PyPI, and npm ecosystems;
• Built supporting infrastructure including TypeScript SDKs, Next.js marketplace systems, APIs, and developer integrations;
• Focused on verifiable AI, cryptographic integrity, GPU optimization, and scalable proof-generation systems for production-grade AI workloads;
• Represented the project publicly through technical demos, open-source ecosystem development, and speaking engagements, including zkSummit 14 in Rome.

A blockchain and AI infrastructure platform focused on scalable digital economies, intelligent creator tooling, and real-time marketplace systems. It combined blockchain infrastructure, AI-powered automation, analytics, and distributed backend services to support gaming ecosystems, creator economies, tokenized assets, and large-scale digital marketplaces. The platform leveraged Substrate, Rust, GraphQL, Next.js, and distributed backend services to power high-throughput Web3 infrastructure while integrating AI-driven workflows and automation for creators, operators, and marketplace participants.

• Served as Lead Protocol Engineer responsible for protocol architecture, intelligent marketplace systems, AI-driven tooling, and distributed backend infrastructure;
• Designed and developed scalable marketplace infrastructure using Rust, Substrate, TypeScript, GraphQL, and Next.js;
• Built AI-oriented backend systems for analytics, marketplace intelligence, behavioral insights, and automated ecosystem operations;
• Architected real-time event-processing pipelines and data orchestration systems for blockchain activity, digital assets, and marketplace interactions;
• Developed intelligent dashboards and analytics platforms capable of processing large-scale ecosystem and transaction data;
• Built automation systems and AI-assisted tooling for Discord and Telegram integrations, improving operational workflows and community engagement;
• Worked on recommendation-style systems and data-layer infrastructure supporting creator discovery, marketplace activity analysis, and ecosystem optimization;
• Designed scalable APIs and developer tooling enabling integration between AI systems, marketplace services, and blockchain infrastructure;
• Contributed to distributed systems architecture handling high-throughput transactions, asynchronous processing, and real-time marketplace synchronization;
• Collaborated across AI engineering, blockchain infrastructure, frontend systems, and product strategy to deliver production-grade Web3 and intelligent platform tooling;
• Focused heavily on system scalability, low-latency architecture, automation, and intelligent operational tooling across the platform ecosystem.

An AI simulation platform for bridging virtual and physical worlds through digital twins, synthetic data generation, and high-fidelity 3D environments for training AI systems. It was designed to help models learn spatial reasoning, environmental understanding, motion prediction, and physical interaction through realistic simulation instead of relying only on real-world data collection.

• Led ML and simulation systems behind the platform;
• Built simulation infrastructure using Unreal Engine 5, PyTorch, Rust, NVIDIA tooling, and Kubernetes;
• Developed synthetic-data and computer-vision pipelines for training AI models in simulated environments;
• Worked on systems that connected virtual simulation, real-world perception, and AI model training;
• Built customer-facing web tooling using React, TypeScript, and Node.js job orchestration on AWS Batch;
• Developed computer-vision pipelines using PyTorch, OpenCV, and TensorRT for NVIDIA edge devices;
• Contributed to the enterprise synthetic-data product from R&D stage to $1M+ ARR;
• Worked across AI simulation, synthetic data, 3D systems, GPU inference, edge deployment, and production ML infrastructure.

A global cryptocurrency exchange platform focused on high-availability trading infrastructure, liquidity systems, real-time market data, and secure digital asset exchange. The work centered on low-latency backend systems for exchange operators, including matching-engine infrastructure, liquidity protocols, WebSocket data distribution, and fault-tolerant order book systems.

• Built matching engines, liquidity protocols, and HFT infrastructure for tier-one cryptocurrency exchange operators;
• Developed backend services using Node.js, TypeScript, C++, Redis, and WebSocket systems;
• Implemented active-active deployment architecture for resilient exchange operations;
• Built custom WebSocket fan-out systems for real-time market data and trading events;
• Built fault-tolerant order book systems with strict latency budgets;
• Engineered hot-path state management using Redis and performance-sensitive backend components;
• Contributed to low-latency trading infrastructure with strict performance and reliability requirements;
• Collaborated on infrastructure serving high-throughput crypto trading, liquidity management, and market-data systems;
• Helped deliver production systems where latency, uptime, and correctness were critical to financial operations.

An AI-powered 3D retail metaverse platform for enterprise brand experiences, immersive commerce, real-time multiplayer environments, and XR-enabled product visualization. It combined AI, 3D engines, web technology, and enterprise deployment workflows to help brands create interactive digital retail spaces across web, mobile, and XR devices.

• Delivered full-stack engineering for the management console and 3D runtimes;
• Built WebGL clients and real-time multiplayer networking;
• Shipped XR integrations for Meta Quest and Rift;
• Led LATAM regional expansion, localization, and on-prem deployments;
• Worked across product, engineering, regional business development, and immersive experience delivery for enterprise customers;
• Built management consoles and customer-facing tools using React and Node.js;
• Developed 3D runtime experiences using Unity, Unreal Engine, WebGL, and real-time multiplayer networking;
• Supported enterprise deployments, localization, and on-premise implementations across LATAM;
• Helped scale the AI-powered 3D retail metaverse platform to $15M ARR.