Client background

Founded in 2003 and backed since 2024 by EQT, one of Europe's largest private equity firms, Avetta is building the largest global community of hiring clients and suppliers that are Ready to Work. Its unified platform streamlines compliance, prequalification, safety and performance benchmarking in a single, integrated experience.

Trusted by over 130,000 businesses across more than 120 countries, Avetta blends AI-driven insights and human expertise to close risk gaps and strengthen supplier reliability so projects start on time, risks are managed proactively, and operations scale with certainty.

Operational challenge

Verifying supplier documentation at scale is one of the most operationally intensive parts of the business. As submission volumes increased across document types, languages, and regions, the legacy approach became harder to sustain efficiently. Scaling capacity depended heavily on specialist review, peak periods increased operational pressure and cost, and multilingual coverage created additional complexity.

The result was a verification operation that was expensive, challenging to scale, and increasingly difficult to keep aligned with Avetta's growth trajectory.

Output by Tintash

Tintash provided Forward Deployed Engineers (FDEs) who worked closely with Avetta's product and engineering teams to bring the company's vision for the Batch Verification Service (BVS) to life. Together, the teams built an end-to-end AI-powered document processing pipeline that removes the manual extraction and rule-checking burden from reviewers while keeping a human in the loop for exceptions and edge cases.

The solution was designed to:

• Streamline document processing across a wide range of submission types.
• Support multilingual verification workflows across regions.
• Improve consistency and transparency in verification outcomes.
• Maintain human review for exceptions and edge cases.
• Preserve auditability within the broader operating environment.

Capabilities and business impact

What the Batch Verification Service is capable of:

Client background and business problem

Design reviews were slow, inconsistent, and impossible to scale

Picture a typical design review before onBeacon. A product team opens a Figma file, shares a screen, and spends an hour going back and forth. One designer flags the visual hierarchy. Another disagrees. A third raises an accessibility concern that contradicts the first. The meeting ends with a list of feedback items, most of them subjective, some of them in conflict with each other. Then someone has to turn that into actual design changes. That process takes another week.

For fast-moving product teams, this was simply the norm. The quality of a design review depended entirely on who was in the room and how much time they had. There was no way to make the output consistent across reviewers, no mechanism to grow throughput without adding people, and no path from feedback to implementation that did not involve another round of meetings.

onBeacon set out to change that. The goal was a platform that could analyze any UI design in real time, surface structured evidence-backed findings, and go one step further by generating concrete prototypes showing teams exactly what a recommended change would look like applied to their actual work. Not abstract feedback. Something a designer could act on in the same session. They needed a technical partner who could own the architecture and build something production-ready from scratch.

How did Tintash deliver?

A production multi-agent architecture of 30-40 specialized agents, with Claude handling the most reasoning-intensive workloads

Tintash owned the system architecture, AI orchestration layer, model evaluation framework, backend implementation, and deployment infrastructure, building the platform from the ground up. The platform has been in continuous production use for two years, evolving through multiple generations of AI models. It has active users across a Figma plugin, Web Dashboard, and a managed service where onBeacon’s designers moderate AI-generated analysis for enterprise clients. Human reviewers remain in the loop for enterprise engagements, validating AI-generated recommendations before delivery to customers.

A multi-agent system at the foundation

The platform routes each design through a production multi-agent system composed of 30–40 specialized agents running in parallel. Each agent evaluates a specific dimension of quality: layout, typography, accessibility, CTA clarity, brand alignment, and other categories. This decomposition allows individual dimensions to be independently benchmarked, tuned, and upgraded without affecting the rest of the system. A resolution agent then consolidates findings across all agents into a single coherent output, handling cross-agent conflicts the way a senior reviewer would.

The platform is multi-model by design. Each agent is benchmarked independently and model selection follows what performs best for that specific task. Models are evaluated on latency, cost, visual understanding, reasoning quality, and actionability of outputs before assignment to production workloads. Visual inspection agents, reading color values, layout structure, and typography, run on GPT-4o where structured pattern recognition is the primary requirement. For tasks that demand both image understanding and complex multi-step reasoning at the same time, the platform uses Claude’s models.

Where Claude Opus powers the platform's highest-value reasoning workloads

Claude Opus powers two of the platform's highest-value reasoning workloads: UX Critique and Prototype Generation. The selection came out of direct model comparison testing. Claude was selected because it consistently outperformed alternatives on traceable reasoning, grounding recommendations in established UX principles, and generating actionable outputs that designers could implement without additional interpretation.

Most models evaluated could identify surface-level design issues but struggled when the task required grounding analysis in established UX, cognitive psychology, and human-computer interaction principles. Getting to specific, traceable reasoning rather than generic observations turned out to be the hard part.

Claude Opus performed significantly better on that requirement. The analysis it produced was grounded in established principles, the reasoning was traceable, and the output was structured, evidence-backed, and directly actionable.

UX Critique

The UX critique layer runs on Claude Opus. It evaluates user flows and design decisions against established UX, cognitive psychology, and human-computer interaction principles, holding context across a full design system rather than responding to isolated elements. The output is the kind of structured, reasoned critique that previously required a senior designer and a synchronous review session. It is now available on demand, consistently, at any scale.

Prototype Generation

After surfacing findings, the platform uses Claude Opus to generate design variants and visual prototypes showing teams exactly what a recommended change would look like applied to their actual design. A finding about visual hierarchy is easy to dismiss. A Claude-generated prototype showing the corrected layout, in context, applied to the team's actual work is not.

For simpler pipeline tasks such as restructuring outputs and formatting analysis for the front end, the platform uses Claude Sonnet, where the task does not require the same reasoning depth and a lighter model performs well. The team’s approach to model selection is to find the model best suited to the job and re-evaluate as new versions ship. As Anthropic models have raised the ceiling on reasoning and image understanding, Claude’s role in the pipeline has grown. The roadmap includes expanding Claude’s role across additional analytically demanding workloads within the system.

Impact for client

Two years in production and in use by design teams at some of the world's largest organizations

onBeacon has been in sustained production for two years. Design teams at global organizations including ByteDance, Salesforce, Siemens, Dubai Govtech Agency, and Singapore Govtech use the platform. AI recommendations have been implemented in live product roadmaps and active sprints.

• 19,000+ product/design users acquired organically
• 30% month-over-month growth in signups
• Time from design insight to implementation reduced from weeks to minutes
• $1.3M pre-seed raised, with the platform as the core proof point
• AI-generated design recommendations implemented in live product roadmaps and sprints

The design review process that once required a room full of people, a shared screen, and a week of follow-up now runs in minutes. Findings are grounded in principles, prototypes are ready to act on, and the output is consistent regardless of who is doing the reviewing.

Client’s brief background and business problem

This US-incorporated client was one of the largest and most compliant asset exchanges in the country. They were looking to onboard specialists who could help support and speed up their asset integration efforts.

Although this client operated on a remote-first model, there were really very few locations outside of the US, Canada, and Europe from where they allowed their team to work from. However, they were in the hunt of a very specific skill set – expertise in Rosetta API specifications. Left with few team members who fully understood how Rosetta worked, the client engaged Tintash given our deep expertise in this area to help eliminate this resource bottleneck.

How did Tintash deliver?

It took Tintash almost three months to get all the required security clearances and access to the client’s systems and machines. Once we started, our team surprised everyone by delivering a complete Rosetta implementation in a mere three weeks. We also helped enable an asset in the quickest ever time in the client’s entire experience of asset integrations.

If our first rosetta implementation showcased our technical ability, our first integration highlighted our capacity to collaborate with multiple teams and stakeholders while keeping everyone focused on the common goal. Positively surprised by our team’s performance, the client handed over the complete asset integration to us.

Our team did not just go above and beyond to deliver the complete end-to-end integration, we did so while also training the client’s engineers. We also interfaced with at least a dozen asset issuers and helped them resolve issues in their Rosetta implementations for a smoother integration.

The third challenge was validating the Rosetta implementations built by the asset issuers for their respective blockchains while also working closely with them to sort out the issues in those implementations. Rosetta being open source doesn’t dictate a lot of those things to be required. However, since we had worked on a complete end-to-end integration, our team foresaw those issues and got them resolved before the actual integration kicked off. This helped further speed up the integration process.

At that point, the client’s executive and product team started looking up to us to drive the optimization efforts in the asset integration pipeline. Our team conducted the root cause analysis and identified Rosetta completeness as still being the biggest bottleneck.

For the next couple of months, we worked closely with the product team to further solidify the Rosetta specifications, which were internal to our client, and effectively share them with all the new assets we were validating and integrating.

How did the final product look?

To summarize, our first Rosetta implementation in a mere three weeks was very well received and unlocked opportunities for us to work on the E2E Asset Integration of numerous assets. Our team also worked closely with the asset issuers to provide them a smooth onboarding experience while getting them to deliver on our requirements.

At the time of this writing (Q1 '23), Tintash continued to be the only team contracted by the client to work as an integral part in their asset integration. Our team members were all working on different assets with the client engineers to provide their expertise and help speed up the multiple asset additions.

Another noteworthy contribution was our team’s squad-building efforts for the client. We got the mandate to build a squad in the US and Canada. We not only recruited and trained IC5 and IC4 engineers across North America but also successfully onboarded them on our client platforms and systems to deliver yet more value to the client.

Client's business problem

The immutable nature of deployed smart contracts creates a critical security concern, as post-deployment vulnerabilities are typically irreversible, leading to substantial financial and reputational damage. As the ecosystem matures, escalating complexity increases the risk of exploits.

Traditional auditing processes involve manual review by consultants, which can take weeks or even months, delaying project timelines. Existing static analysis tools, though efficient, frequently fail to detect vulnerabilities stemming from complex business logic. Organizations need an automated, intelligent system that accelerates audits, reduces human error, and provides deeper insights beyond static code scanning.

Clara wanted to solve this problem for the Stacks ecosystem by building a tool that could comprehensively analyze Clarity smart contract logic, provide actionable insights, and integrate seamlessly into existing workflows without disrupting the development speed.

How did Tintash deliver?

Tintash designed Clara, a formal verification engine powered by an AI audit agent that performs deep static and dynamic analysis to detect vulnerabilities beyond surface-level scanning. The development team faced several challenges, including the need to train AI models specifically for Clarity, the smart contract language for the Stacks blockchain. Unlike Solidity-focused tools, Clara had to understand and adapt to Clarity's unique nature - Turing-incomplete and decidable among other characteristics - to avoid false positives and improve detection accuracy.

Team Achievements:

• Developed a multi-agent AI audit system capable of modeling vulnerabilities based on real-world attack patterns.
• Overcame bias in existing LLMs by finetuning AI models with Clarity-focused datasets to ensure accurate vulnerability detection.
• Implemented formal verification techniques that mathematically prove the correctness of smart contracts, providing stronger security assurances compared to traditional fuzzing techniques.
• Designed an intuitive workflow that allows developers to upload contracts via GitHub or direct file submission and receive detailed security assessments within minutes.
• Clara delivers a fully automated, end-to-end contract audit - no manual steps required. Simply upload a smart contract or connect your code repository, and the platform generates a comprehensive final audit report highlighting vulnerabilities and compliance insights.

The biggest challenge was ensuring that Clara could go beyond static code analysis to detect business logic errors and systematically uncover security vulnerabilities, which often remain undetected by conventional tools. Our team developed an AI-powered machine-aided remediation engine that not only flags security vulnerabilities but also suggests fixes based on best practices, reducing developers' burden in securing their code.

How did the final product look?

Clara is a web-based AI audit platform that provides a streamlined, developer-friendly interface for smart contract security assessments. The dashboard presents key security metrics such as threat scores and vulnerability categorization, enabling quick risk assessment. The tool generates detailed vulnerability reports, offering precise explanations and mitigation strategies. Clara's formal verification outputs validate smart contract correctness with machine-assisted reasoning through mathematical theorem proofing, ensuring higher reliability. Its integration capabilities allow developers to use it within future GitHub Actions, CLI environments, and Visual Studio Code extensions, making security auditing a seamless part of their workflow.

The fact that the audit engine is continuously being updated, based on the industry security benchmarks, ensures the audit engine stays current, identifying both known and emerging vulnerabilities. This allows Clara clients to deploy contracts faster with enhanced security confidence. By enabling developers to catch security flaws early, accelerate smart contract deployment, and ensure a higher level of trust and security in decentralized applications, Clara is redefining how blockchain projects approach security audits.

Client background and business problem

Above Data was an early-stage signal infrastructure platform with a strong thesis: marketing and CPG brand teams should be able to build, refine, and activate signal-based audiences from live, deterministic signal without depending on data engineers, SQL queries, or week-long analyst turnarounds.

To validate that thesis with investors and prospective enterprise customers, they needed a working MVP they could put in front of a fundraising audience. Not a clickable prototype or a deck, but a real, multi-tenant product that a live user could drive end-to-end against real consumer data.

The MVP had to be:

• Polished enough to anchor the fundraising story
• Technically credible enough to survive due diligence
• Architecturally sound enough to support the company through its first enterprise customers post-raise

Their target customers, marketing teams at retailers and CPG brands, had access to valuable first-party and panel data, but no intuitive way to interact with it.

Building a targeted audience typically required:

• Filing a ticket with a data engineer
• Waiting for a SQL-driven extract
• Iterating over email
• Manually loading the resulting list into ad platforms

The cycle could take days, limiting campaign agility and restricting the product to a small group of analysts. Above Data approached Tintash with a clear ambition: build a fundraising-ready MVP that rebuilt the audience-building experience as a conversational, AI-native product any marketer could use.

Key discovery questions included:

• How do users describe an audience in plain English while still ending up with a precise, deterministic signal definition that ad platforms can ingest?
• How do we present millions of consumer records in a way that is digestible, interactive, and actually guides the user toward a high-quality audience?
• How do we keep the AI co-pilot grounded in real data and not hallucinated numbers?
• How do we ship this as a true multi-tenant SaaS product with enterprise-grade authentication from day one?
• How do we close the loop by delivering audiences directly to Facebook, Google, and other destinations without CSV exports?
• How do we do all of this on an investor-driven timeline with the polish and reliability needed for live fundraising demos?

How did Tintash deliver?

Tintash designed and built the Above Data MVP end-to-end, creating a fundraising-ready, multi-tenant platform that turned analyst-led signal-based audience creation into a self-serve product for enterprise marketers.

Tintash owned the full product stack:

• Product design and Figma prototyping
• Next.js frontend development
• Agentic AI product development
• Python AI agent backend
• Data and delivery integrations
• Google Cloud Platform infrastructure

Rather than splitting the project across multiple vendors, Above Data partnered with a single team that owned the product from the first design prototype to production deployment.

3. Click-to-Refine Audience Ingredients

Every AI-generated chart is interactive.

Clicking a recency band, brand, or spending range:

• Adds the selection to a structured Audience Ingredients panel
• Updates the live consumer count in real time
• Creates a deterministic and auditable audience definition

This gave marketers complete visibility into exactly how an audience was being built.

4. Multi-Tenant SaaS Architecture

Above Data needed a credible path to enterprise revenue.

Tintash built:

• Subdomain-based tenant routing
• Organization-aware authentication
• Secure multi-tenant data architecture
• Enterprise-grade cloud infrastructure

Tintash also owned:

• Cloud environment setup
• Identity and access controls
• Storage and networking architecture
• Deployment pipelines
• Environment templates

New enterprise tenants can now be onboarded through configuration instead of engineering cycles.

6. Four-Step Delivery Wizard

Building an audience was only half the workflow. Activation was the other half.

Tintash designed and built a guided delivery flow:

Select Audience → Destination → Parameters → Preview

Audiences can now be pushed directly to:

• Facebook Ads
• Google Ads
• Other destinations

With:

• Audience summaries
• Match-rate visibility
• Delivery validation

This removed the need for CSV exports and manual uploads.

What the final product looked like

The final MVP became a polished, AI-native marketing product that Above Data can demo live during investor and enterprise conversations.

The platform delivers:

• A personalized dashboard experience
• Draft audiences that can be resumed
• Audience insights and forecasting
• Upcoming delivery schedules
• End-to-end audience creation and activation
• Per-tenant enterprise environments that can be provisioned without re-architecting the platform

Impact for client