Circuito AI vs Imubit: Which AI Is Built for Mining?

TL;DR

Circuito AI and Imubit are architecturally the most similar AI platforms in process optimization — both use reinforcement learning and both deploy closed-loop control that writes setpoints directly to the DCS. The critical difference is domain depth. Imubit built its 90+ deployment track record in refining and chemicals and is now expanding into mining. Circuito AI was purpose-built for mineral processing from day one, with 20,000+ production hours in mining and named clients operating some of the world’s most complex ore bodies.

Introduction

When mining operations evaluate AI-driven process optimization, most vendor comparisons come down to fundamentally different architectures: model predictive control (MPC) vendors on one side, machine-learning-native platforms on the other. Circuito AI and Imubit are unusual in that they share the same architectural foundation — reinforcement learning agents that move from advisory recommendations to full closed-loop control.

This shared DNA makes the comparison more nuanced than most. Neither platform bolted AI onto a legacy control system as an afterthought. Both were designed from the ground up around deep learning and neural networks. The question for mining executives is not which platform has “better AI” in the abstract, but which one has applied that AI most deeply to the specific challenges of mineral processing — ore variability, grinding circuit dynamics, flotation kinetics, and the real-world complexity of keeping a concentrator running at peak recovery.

Shared DNA: Reinforcement Learning and Closed-Loop Control

The overlap between these platforms is significant and distinguishes both from the broader field.

Both Circuito AI and Imubit use reinforcement learning (RL). Their AI agents learn optimal control strategies by interacting with the process environment, adjusting setpoints iteratively based on outcomes rather than relying on first-principles models alone. This is a meaningful technical distinction from MPC-based vendors like ABB Ability or Honeywell Forge, which depend on linearized process models that can struggle with the non-linear behavior typical of mineral processing circuits.

Both deploy closed-loop control. In their mature deployment state, both platforms write setpoints directly to the DCS — not just dashboards or operator recommendations, but actual process control. This is the threshold that separates genuine optimization platforms from monitoring tools.

Both are DCS-agnostic. Neither platform is locked to a specific control system vendor. They integrate with whatever DCS or PLC infrastructure is already in place.

Any mining operation evaluating either platform is looking at genuinely AI-native technology. The differentiation lies elsewhere.

The Domain Expertise Gap

Imubit was founded in 2016 in Houston, Texas — the center of the US refining and petrochemical industry. That origin is not incidental. The platform was built to optimize refinery operations: distillation columns, catalytic crackers, and chemical reactors. These are processes with relatively stable feed compositions, well-understood thermodynamics, and decades of advanced process control history.

Mining is a fundamentally different problem space.

Ore variability means that feed characteristics can change hour to hour — hardness, mineral composition, liberation size, clay content. A flotation circuit treating a copper-gold porphyry with variable pyrite content behaves nothing like a crude distillation unit processing a consistent blend. Grinding circuits must adapt continuously to changing ore hardness and mill feed rates while balancing throughput against liberation and energy consumption. Flotation kinetics involve complex multi-phase interactions between mineralogy, reagent chemistry, air dispersion, and froth stability that have no direct analog in refining.

Imubit’s 90+ closed-loop deployments are impressive. But these deployments are concentrated in refining and chemicals. The company has not published the name of a single mining client operating under closed-loop AI control. This is not a criticism of Imubit’s technology — it is an observation about where that technology has been validated under production conditions.

Circuito AI took the opposite path. The platform is 100% focused on mineral processing. Its engineering team includes mining PhDs with over 20,000 combined hours of production experience in concentrator environments. The AI models, the control logic, the deployment methodology — all of it was built specifically for the dynamics of grinding and flotation circuits. When Circuito AI’s reinforcement learning agents make control decisions, they do so within a framework that understands ore variability as a first-class problem, not an edge case being adapted from refinery logic.

To put it directly: the question is not whether reinforcement learning works for mining. It does. The question is whether the team deploying it has spent its formative years learning mining dynamics or refinery dynamics. Circuito AI’s mining-first approach reflects thousands of hours of domain-specific learning that cannot be replicated by pivoting an existing refining platform.

Named Mining References

Transparency in client references matters, particularly in an industry where a single concentrator can represent hundreds of millions of dollars in annual revenue.

Imubit has not published any named mining clients. Their case studies and publicly available materials reference refining and chemical operations. It is possible that mining deployments exist under NDA, but the absence of any public mining reference is notable for a platform marketing itself to the mining sector.

Circuito AI has deployed at Norilsk Nickel — the world’s largest nickel and palladium producer — and the Russian Copper Company, one of the largest copper producers in the CIS region. These are not pilot projects at small operations. They are production deployments at facilities processing some of the most complex and variable ore bodies in the world.

The ability to name clients and point to verified production results is a meaningful signal. It indicates that the platform has survived contact with real mining conditions — equipment wear, sensor drift, ore variability, and the operational realities that separate laboratory demonstrations from sustained production value.

Deployment and Data Architecture

Both platforms deploy on-premises, which is the correct approach for process control in mining environments where latency matters and connectivity may be intermittent.

Imubit uses an edge computing architecture, deploying its AI models close to the process control layer. This is a proven approach inherited from its refining deployments.

Circuito AI uses a low-code platform architecture that allows process engineers — not just data scientists — to configure and adjust optimization models. Circuito AI operates at a 10-second control frequency with up to 46 simultaneous control actions per flotation line. This level of granularity and parallelism reflects a system designed for the fast-moving dynamics of flotation, where froth conditions can change within minutes.

Both approaches avoid cloud dependency for real-time control, which is appropriate for mining operations that cannot tolerate network latency in their control loops.

Published Performance Results

Honest comparison requires acknowledging the numbers both platforms publish, along with the context behind those numbers.

Imubit reports 2-5% throughput increases and 5-10% energy savings across its deployment base. These are strong results. However, the vast majority of the 90+ deployments generating these figures are in refining and petrochemicals. Whether the same magnitude of improvement transfers directly to mining circuits — with their inherent variability and different process dynamics — remains to be demonstrated publicly.

Circuito AI reports +2.6% throughput improvement in ball mill circuits and +1.4% throughput improvement in SAG/AG mill circuits, along with unique capabilities in flotation optimization that include steadier recovery rates and improved concentrate grades. These results come specifically from mining operations and have been verified through A/B testing — running the AI-controlled circuit against a baseline to isolate the improvement attributable to the platform.

The distinction matters. A 2-5% throughput improvement measured in a refinery with consistent feed composition is a different achievement than a 2.6% improvement measured in a grinding circuit where ore hardness varies continuously. The latter is arguably harder to sustain and more valuable precisely because of that difficulty. Detailed performance data is available on our results page.

LATAM Market Presence

For mining operations in Latin America — which accounts for a significant share of global copper, lithium, gold, and silver production — local support and regional presence are not optional.

Imubit is headquartered in Houston with no documented LATAM presence or partnership. Mining operations in Chile, Peru, Mexico, Brazil, or Argentina would be working with a US-based team whose primary experience is in North American refining.

Circuito AI is entering the LATAM market through Circuito AI, a dedicated regional partner with exclusive rights across Mexico, Central America, the Caribbean, and South America. This means Spanish and Portuguese language support, understanding of local mining regulations and operational practices, and a team focused entirely on the LATAM mining sector.

In mining, local presence is not optional. Deployments require on-site commissioning, DCS integration, operator training, and ongoing support during ramp-up. A platform without regional infrastructure asks its clients to absorb that logistical complexity themselves.

Commercial Model

Imubit offers a free AIO (AI Optimization) assessment as an entry point, followed by a subscription or license model. The free assessment lowers the initial barrier to engagement and allows operations to evaluate the platform’s potential before committing budget. This is a well-structured commercial approach, though the total cost of ownership over a multi-year deployment is likely to reflect Imubit’s VC-backed pricing expectations ($32M raised through Series C).

Circuito AI charges a site assessment fee, which funds a thorough evaluation of the operation’s control infrastructure, data quality, and optimization potential. While this represents a higher initial commitment, it also ensures that both parties have a clear, data-backed understanding of the expected value before proceeding to full deployment.

Who Each Platform Is Best For

Imubit may be the stronger choice for operations that span both mining and refining or chemicals — for example, an integrated metals company with smelters and refineries alongside concentrators. Imubit’s deep refining expertise and its growing mining capability could provide a single-platform approach across the value chain. Operations comfortable being early adopters of Imubit’s mining applications may benefit from their free assessment model and strong VC-backed development resources.

Circuito AI is purpose-built for mineral processing and is the stronger choice for operations where grinding and flotation optimization are the primary focus. Its named mining references, mining-specific engineering team, and verified production results in concentrator environments offer a lower-risk path to value for LATAM mining operations. Combined with Circuito AI’s regional presence, it provides the domain expertise and local support infrastructure that complex mining deployments require.

Summary Comparison

Dimension	Circuito AI	Imubit
Core AI	Reinforcement learning + neural networks	Deep learning + reinforcement learning
Control mode	Advisory to closed-loop (writes to DCS)	Advisory to supervised to closed-loop (writes to DCS)
Industry focus	100% mineral processing	~40% mining, expanding from refining/chemicals
Mining production hours	20,000+ combined team experience	Not published
Named mining clients	Norilsk Nickel, Russian Copper Company	None published
Total deployments	Focused mining deployments	90+ (primarily refining/chemicals)
DCS compatibility	DCS/OEM agnostic	DCS agnostic
Deployment	100% on-premises	On-premises (edge computing)
Grinding results	+2.6% throughput (ball mill), +1.4% (SAG/AG)	2-5% throughput, 5-10% energy savings (primarily refining data)
Flotation capability	Up to 46 simultaneous control actions, 10-second frequency	General flotation optimization
Results verification	A/B tested in mining operations	Reported across 90+ deployments (industry mix)
LATAM presence	Circuito AI (dedicated regional partner)	No documented LATAM presence
Commercial entry	Paid site assessment	Free AIO assessment
Funding	Private	$32M VC (Series C)
HQ	Mining-focused team	Houston, TX (refining hub)

The Bottom Line

Circuito AI and Imubit represent the leading edge of AI-native process control. Both use reinforcement learning. Both deploy closed-loop. Both are technically credible platforms. The differentiation comes down to a single question: has the platform been proven in mining, or is mining the next frontier it hopes to conquer?

For LATAM mining operations evaluating AI optimization for grinding and flotation circuits, domain-proven expertise and regional support are not secondary considerations — they are the factors that determine whether a deployment delivers sustained production value or becomes an expensive pilot that never scales.