Aztecai GmbH connects decentralized AI infrastructure with practical AI assistants. From Bittensor operations to Teki, we build around measurable performance and human control. Our next focus: agentic AI security and governance in the EU.
Aztecai GmbH is small, deliberate, and obsessive about quality. The name carries a Mexican founder's admiration for the Aztec engineers — same discipline, new tools.
Bittensor is a decentralized AI network — a protocol where independent operators compete to provide the best machine intelligence, and are paid for it in TAO, the network's native token. Think of it as an open, permissionless marketplace for AI: anyone can plug in compute and models, anyone can consume the output.
The network is divided into subnets, each a self-contained AI market focused on one task — text generation, image synthesis, quantum simulation, drone autopilots, decentralized inference, and dozens more.
Miners are operators who run the actual AI workload for a subnet — serving inference, training models, simulating circuits, generating images. They commit real compute (often GPU fleets) and compete on quality.
Validators continuously test miners on benchmark tasks, score the results, and publish a quality ranking on-chain. The better a miner performs, the higher their weight in the subnet's ledger.
Miners receive subnet alpha rewards according to the network's emission rules and validator-assigned weights. Scoring and allocation vary by subnet; strong benchmark performance alone does not guarantee revenue.
Since the dTAO upgrade (Feb 13, 2025), each subnet issues its own alpha token whose price floats against TAO based on supply and demand. Miners receive alpha for their work and can swap it back to TAO on-chain. TAO itself trades on major exchanges — connecting the network's machine-intelligence output to real capital markets. Token prices, demand, competition and infrastructure costs all affect operating outcomes.
We operate decentralized AI infrastructure and develop, test and optimize miners across specialized Bittensor subnets. Our work connects model serving, adversarial robustness, computer vision and autonomous agents—with measured quality, latency and operating economics guiding each decision.
Our portfolio preserves the subnets we mined and worked on during 2025 and 2026—even when a project has since closed, changed name or left the network. Historical participation is not a claim of current mining activity. Research-only work is identified separately.
Subnets are counted once across the historical portfolio, including retired projects; research-only tracks and root staking are excluded.
Operating GPU infrastructure for serverless model serving. Our work centers on reliable capacity, model placement, utilization and the relationship between performance and sustainable operating cost.
Explore Chutes →Providing GPU capacity and monitoring rental utilization, availability and provider economics. We track rental income separately from subnet rewards to understand the real operating performance.
Explore Lium →Our historical subnet footprint, retained from AztecAI's 2025 company portfolio. We preserve these entries whether or not the original projects still exist. Names shown are historical; subnet IDs can be reassigned to unrelated projects.
Where the original record only identified a subnet number, we retain that number rather than substitute today's project name. Root-network staking and validator rewards are not included as mined subnets.
Our 2026 portfolio includes mining operations and development across Targon, Perturb, Score Vision, MIID, Leadpoet, REAX, Swarm and Poker44, alongside Chutes and Lium above. Engy and Harnyx are research tracks; Ditto covers agent-memory development and evaluation. These are different stages of work, not a list of currently active miners.
Read the actual evaluation code. Establish a reproducible baseline. Compare alternatives on the same tasks. Review security and failure modes. Promote changes only after testing—and monitor the result. We separate benchmark scores from real-world performance, and network rewards from revenue, cost and cash flow.
Bittensor pushes us to measure how AI systems perform under real constraints. Teki brings that engineering mindset to a different problem: making a useful assistant part of everyday life and work.
Our direction is practical: assistants with a defined purpose, continuity across conversations and access to selected tools—not unrestricted autonomy. Which capabilities are enabled depends on the assistant, its permissions and the deployment.
An agent does more than generate text. It can access data, invoke tools and act on behalf of people. Those capabilities need clear boundaries, accountable ownership and evidence that controls work.
Over the coming months, AztecAI will develop this practice for organizations operating in the European Union—bringing our cybersecurity background together with hands-on agent engineering.
Discuss a pilot →This is a developing engineering and governance practice—not a certification, legal opinion or guarantee of compliance. Obligations depend on the system, its intended use and the organization's role.
Four core practices. Tailored to your team, your data, and your roadmap.
Custom AI software tailored to your business — from RAG pipelines to bespoke ML services and inference APIs.
Tailored AI agents for small businesses and entrepreneurs — built around your daily operations to automate repetitive work and reclaim hours every week.
Robust, observable, production-grade AI. We build evals, monitoring, and guardrails into every system from day one.
Active contributors to the Bittensor ecosystem — designing, validating, and running subnets that push decentralized intelligence forward.
Jaime spent close to a decade building cybersecurity into safety-critical automotive systems — Project Security & Privacy Manager at Continental, cybersecurity consultant on ZF Friedrichshafen powertrain and Continental BCM/Infotainment programs. He is certified in applying ISO/SAE 21434 and UNECE WP.29/R155 to safety-critical products — and that discipline now shapes how AztecAI builds AI.
AztecAI connects three areas: decentralized AI engineering through Bittensor, practical assistants through Teki, and a developing agentic AI security and EU governance practice. Experience operating and testing real systems informs how we approach permissions, evidence and human oversight.
He combines technical depth with direct business communication — earlier sales engineering with National Instruments and Schneider Electric, sharpened by an MBA from Warwick Business School. He works directly with every client. No middle layers, no junior handoffs.
Tell us about it. We respond within 48 hours with a clear assessment — feasibility, scope, and what it would take to ship.