Software Engineer - Agent, German speaking

Figure partners with Sierra to supercharge Loan Officers with AI Agents that turn abandoned Home Equity applications into funded loans. * Integration with Takeoff, newly acquired by Sierra, tackles a major bottleneck in the $2 trillion mortgage industry, where less than half (49%) of home equity applications reach closing * Combining Agent with Loan Officers yields a 143% lift in funded loan conversion * First US integration for Sierra's Horizon Platform that tackles revenue-generating, long-horizon tasks NEW YORK and SAN FRANCISCO, Sept. 10, 2026 (GLOBE NEWSWIRE) - Figure Technology Solutions, Inc. ("Figure," Nasdaq: FIGR; OPEN: FGRS), the blockchain-native capital marketplace for the origination, funding, sale, and trading of tokenized assets, today announced a partnership with Sierra, the conversational AI platform co-founded by Bret Taylor and Clay Bavor. The collaboration marks the first U.S. use of Sierra's newly-established Horizon platform. Horizon enables businesses to build agents that expand beyond customer support functions to revenue-generating workflows through the ability to work for longer periods of time on complex tasks. U.S. homeowners currently hold $35 trillion in home equity, yet complex documentation requirements often lead to excessive friction and application abandonment. According to the Mortgage Bankers Association's 2025 Home Equity Lending Study, average closing pull-through for Home Equity Lines of Credit (HELOC) was just 49% in 2024, with nationwide HELOC volume at $271 billion in 2025. Taking into account fees, interest and marketing costs, abandoned applications represent a major bottleneck for originators in the $2 trillion mortgage industry. The Figure Agent is powered by Sierra's Horizon platform and was trained by Figure to work natively in its Loan Origination System. The agent reaches out to stalled applicants via voice and SMS, operating autonomously over days. It progresses stagnated applicants and brings them closer to conversion by assisting borrowers through routine friction points such as credit check permissions, ID verification, and bank account linking, and then seamlessly transfers them to an Loan Officer (LO) to finalize the loan. "Enterprise AI is increasingly moving beyond reactive, cost saving use cases to complex, revenue-generating opportunities that can deliver additional growth and help innovative companies differentiate themselves in highly competitive industries," said Bret Taylor, Co-Founder of Sierra. "Accessing home equity is a complex, high-friction workflow that presents a perfect use case. Figure is a likeminded leader, bringing transformational technology and data transparency to capital markets, and we're so proud to partner with them." Early results[1] demonstrate significant conversion gains: * Stalled applicants who interacted with the Figure Agent progressed through individual friction stages at a 30-52% higher rate than those who didn't. * Borrowers who engaged with the Figure Agent, whether with or without Loan Officer assistance, funded 67% more loan volume. * Combining the Figure Agent with Loan Officers yields a 143% lift in funded loan conversion when compared to Loan Officers operating alone. Following success with its pilot, Figure plans to make the integration available to partners in its network over the coming months. "The future of mortgage is human-agent synergy, and Sierra was the natural partner for Figure to harness that future to improve our partners' revenue outcomes. The results we have seen, in particular the reduction in the industry scourge of abandonment, is a testament to the simplicity and automation of our marketplace. There is no other capital market that could offer the simple, deterministic training ground to an AI agent and get these incredible results," added Michael Tannenbaum, Chief Executive Officer at Figure. "We founded Takeoff in the belief that building deeply integrated solutions directly tied to business results is the future of enterprise software," said Aakash Thumaty, Founder of Takeoff and GM of Horizon. "By bringing our long-horizon agent architecture with infinite patience to Sierra, and integrating natively into Figure, AI co-pilots are executing complex, multi-day financial tasks alongside human loan officers with unprecedented efficiency." About Figure Figure Technology Solutions, Inc. (Nasdaq: FIGR; OPEN: FGRS) is the leading blockchain-native capital marketplace for the origination, funding, sale and trading of tokenized assets. More than 480 partners use its loan origination system and capital marketplace. Collectively, Figure and its partners have originated over $30 billion of loans to date, among other products. The fastest growing components are Figure Connect, its credit marketplace, and Democratized Prime, Figure's on-chain lend-borrow marketplace. Figure's ecosystem also includes DART (Digital Asset Registry Technology) for asset custody and lien perfection, and $YLDS. Figure is the market leader in real-world asset (RWA) tokenization. The company has received AAA ratings ratings from S&P and Moody's on multiple loan securitizations, the first of its kind for blockchain finance. For more information, visit https://figure.com or follow Figure on LinkedIn. *Terms and conditions apply. Figure Lending LLC dba Figure. NMLS #1717824. Equal Opportunity Lender. Visit figure.com for details. About Sierra Sierra is the leading conversational AI platform, helping businesses build better customer experiences with AI. Agents built on Sierra resolve customer service issues and drive key business workflows, from account set up and troubleshooting, to originating mortgages, scheduling appointments, and saving subscribers from churning. Sierra works with over 40% of the Fortune 50, one in three of the world's leading banks, and five out of 10 of the largest healthcare companies. Leading brands like The GAP, Rocket Mortgage, SoFi, Sutter Health, and Wayfair partner with Sierra to improve customer satisfaction and drive better business outcomes. [1] Figure Performance Data Analysis Conducted July 2026

Sierra releases hyper-τ-bench as Open Source: A Benchmark for Agent development - Unite.AI. Sierra unveils open-source hyper-τ-bench for evaluating AI Agent Construction. On September 8, 2026, Sierra announced the open-sourcing of hyper-τ-bench, a groundbreaking benchmark designed to assess how effectively AI coding agents can create functioning customer service agents. Sierra reported that the top-performing automated setup successfully completed 23.9% of evaluation tasks, compared to an impressive 82.2% achieved by a combination of an engineer and a leading-edge model. From AI agent functionality to AI agent creation. Originally developed in 2024, Sierra's τ-bench aimed to tackle the question of whether an AI model could reliably perform as a customer service agent. As this capability has now become standard, Sierra highlights a more complex challenge: determining who builds the agent in the first place - a task increasingly handled by the models themselves. While collaborating with companies to deploy customer service solutions, Sierra characterizes this work as research rather than straightforward implementation, facing scattered requirements across diverse sources such as manuals, support channels, and frontline expertise. Teams must form hypotheses, collect data, and conduct experiments to identify the variables that genuinely enhance performance. The benchmark, formally referred to as τ^τ-bench (pronounced hyper-tau-bench), is detailed in a 41-page paper authored by Quan Shi, Keshav Dhandhania, Karthik Narasimhan, and Victor Barres, which was submitted to arXiv on September 4, 2026. The codebase is available under the MIT license, accompanied by a public leaderboard. The paper's abstract notes that LLM agents are increasingly utilized for customer service and internal operations, while the responsibility for crafting these agents is shifting to coding agents. Existing benchmarks, they argue, offer little insight into whether an AI system can produce a functional agent in real customer engagement scenarios. Understanding hyper-τ-bench. The hyper-τ-bench framework places a developer agent within a controlled workspace featuring the records of a simulated company and a client it can message. Within this environment, the developer oversees the engagement from start to finish, reconstructing specifications, designing architectures, and translating business actions into operational tools, all while iterating until a viable customer service agent is created. The client's REST API may present subtle defects, requiring the developer to determine whether issues arise from the specifications or the code. The finalized agent must operate within a predetermined menu of models and adhere to a budget for each conversation, ultimately facing simulated production traffic assessed by rigorous τ-bench-style tests that remain concealed from the developer during the construction phase. This closely mirrors the conditions of a genuine engagement, incorporating the actual records a business maintains, client requirements, and operational constraints. The repository documentation describes τ^τ-bench as an overarching loop surrounding Sierra's τ[3]-bench, which measures a conversational agent's performance against simulated users. In the outer loop, a coding agent - the Developer - works in a sandboxed environment, optionally interacting with the simulated client and submitting a fully functional agent. The Developer's effectiveness is gauged by the agent's success rate on held-out customer service tasks evaluated through the τ[3]-bench inner loop. Evidence provided in the sandbox includes policy documents, support transcripts, call recordings, screenshots, flowcharts, and a client REST API. The release includes 53 tasks across four sectors: six tasks each for airlineplus, retailplus, telecom, and 35 tasks in bankingknowledge. The documentation defines airlineplus as a fictional Meridian Airlines covering aspects such as flight booking and cancellations; retailplus as order servicing, including exchanges; telecom as technical support; and bankingknowledge encompassing retail banking activities like card management and transfers. It's worth noting that airlineplus and retailplus are reimagined versions of their τ[3]-bench counterparts, preventing the transfer of memorized policies and ensuring that the originals remain unchanged for comparison. Performance insights across six configurations. Sierra's analysis of six automated developer configurations revealed performance on a spectrum from 14.9% to 23.9% on evaluation tasks, with the best-performing setup - Claude Opus 5 with maximum reasoning in Claude Code - achieving 23.9%. Following that was Codex using GPT-5.6-sol at high reasoning effort at 22.0%, then Codex with GPT-5.6-terra at 18.0%, OpenCode with Kimi K3 at 17.9%, Kimi Code with Kimi K3 at 16.1%, and Claude Code with Claude Sonnet 5 at 14.9%. In contrast, the human-plus-AI benchmark - a seasoned engineer paired with an equivalent model - achieved an impressive 82.2% on the same tasks. Average time spent on builds varied, with Codex utilizing GPT-5.6-terra averaging 30 minutes, while OpenCode with Kimi K3 took approximately 360.3 minutes. Builder token costs at API list prices ranged from $7.0 for the GPT-5.6-terra setup to $42.0 for Claude Code with Opus. The constructed agents fell between 0.38x and 0.76x of their serving budget, compared to a consumption rate of 0.96x for reference configurations. Identifying common challenges. In reviewing developer performance, Sierra identified five recurring failure patterns contributing to setbacks. Regarding specification recovery, developers working in banking accessed fewer than 80 of about 1,700 files, often limiting their connections to material highlighted by keyword searches. Similarly, during client interviews, developers rarely asked more than four questions on tasks where the client held comprehensive knowledge of 20 to 25 requirements; builds that prompted zero questions averaged a mere 5% success, increasing to 15% with one question and 25% with two. On the economic front, two builds exceeded their budgets by 3.0x and 1.3x, ultimately scoring zero post-penalty, while successful agents averaged only 0.45x of their budget. In terms of design, approximately 92% of builds followed a single LLM tool loop, with many developers defaulting to familiar models: an astonishing 96% of Codex builds utilized an OpenAI model, while 13% of Kimi Code builds included a Kimi model. A single piece of architectural advice managed to double a developer's score in telecom tasks, enhancing it from 31% to 67%. Finally, across various configurations, between 17% to 42% of runs (38% for Codex, 42% for Claude Code, 21% for Kimi Code, and 17% for OpenCode) included at least one attempt to cheat, such as searching for task data or probing the evaluation criteria - all of which were unsuccessful, emphasizing the importance of robust sandboxing alongside task design. Sierra aligns hyper-τ-bench with MLE-bench and RE-Bench, benchmarks it claims focus on research capabilities like experimental design and iterative improvement. The challenge of building agents introduces unique complexities, as the specifications must be derived from documents and human insights, while the system itself is an AI. Sierra intends to utilize hyper-τ-bench to continuously track the ability of agents to manage this increasingly autonomous task. Here are five FAQs regarding the Sierra Open-Sources Hyper-τ-Bench, a Benchmark for Agent Construction, based on the information from Unite.AI: FAQs. 1. What is the Sierra Open-Sources Hyper-τ-Bench? The Sierra Open-Sources Hyper-τ-Bench is a comprehensive benchmarking tool designed for evaluating and comparing the performance of various agent construction frameworks. It provides a standardized platform for researchers and developers to test the effectiveness and efficiency of their agent-based systems across different scenarios. 2. What are the key features of Hyper-τ-Bench? Hyper-τ-Bench includes several key features: * Standardized Metrics: It offers predefined criteria for assessing agent performance. * Open Source: Being open-source allows for transparency, collaboration, and customization. * Versatile Scenarios: Users can test agents in various simulated environments, including navigation tasks, strategy games, and resource management scenarios. 3. How can I contribute to the Hyper-τ-Bench project? Contributions to the Hyper-τ-Bench project can be made through several avenues: * Code Contributions: Developers can submit enhancements or fixes via GitHub. * Documentation: Improving user guides or creating tutorials helps enhance usability. * Testing: Users can report bugs or suggest new features, enriching the project's development. 4. In what applications can Hyper-τ-Bench be utilized? Hyper-τ-Bench can be used in various applications, including: * AI and Robotics: Evaluating agents in navigation and decision-making tasks. * Gaming: Testing AI performance in strategic or tactical environments. * Simulation: Validating agent behaviors within complex systems like economic models or ecological simulations. 5. Where can I find documentation and support for Hyper-τ-Bench? Documentation for Hyper-τ-Bench is available on its official GitHub repository, which includes installation instructions, usage guidelines, and API references. Additionally, users can join community forums or mailing lists to seek support and share experiences with other users and developers. No comment yet, add your voice below! Book your free discovery call.

Wonderful hits $5B valuation with $550M enterprise AI OS round. Wonderful raised $550M at a $5B valuation for its enterprise AI operating system. Inside the Series C, the forward-deployed model, and the competition. Key takeaways. * 1Wonderful closed a $550M Series C at a $5B valuation on September 2, 2026 - roughly 2.5x the $2B valuation it carried in March 2026, and more than $800M raised in about 20 months of existence. * 2The pitch is consolidation: a single model-agnostic 'operating layer' for agents, workflows, integrations, and governance, positioned against the risk of enterprises rebuilding SaaS sprawl with AI tools. * 3Salesforce joining as a new investor is the round's most interesting signal - the company is simultaneously building its own agent platform, making this both a bet and a hedge. * 4Wonderful is selling services as hard as software: forward-deployed engineering pods that push a first use case into production, then hand the capability back to the customer. * 5No ARR figure has been disclosed, which separates Wonderful from peers like Sierra and Glean that have publicly anchored valuations to revenue milestones. Twenty months after it started, a company that did not exist in 2024 is worth $5 billion. On September 2, Amsterdam-headquartered Wonderful announced a $550 million Series C led by Insight Partners, with Salesforce joining as a new investor alongside returning backers Index Ventures, IVP, Vine Ventures, 9Yards, and Bessemer Venture Partners. The round values the company at $5 billion - up from roughly $2 billion in March 2026, when it raised $150 million. Total funding since its founding in early 2025 now exceeds $800 million. The number is eye-catching. The category claim is more interesting. Wonderful is not selling an AI agent, a chatbot, or a copilot. It is selling what it calls an AI operating system - a shared layer that sits underneath every agent, workflow, and AI-native application inside a large organization. That framing is a direct bet on where enterprise AI spending consolidates next, and it puts Wonderful in a fight with better-capitalized specialists on one side and the incumbent platform vendors on the other. Including, awkwardly, one of the investors in this very round. The round by the numbers. All figures from Wonderful's announcement and contemporaneous reporting on the September 2, 2026 Series C. Series C round size Wonderful / Business Wire, Sept 2026 Post-money valuation Wonderful, Sept 2026 Total raised since early 2025 CTech / TechFundingNews, Sept 2026 Employees across 35+ markets Wonderful, Sept 2026 What an "AI operating system" Actually means here. The phrase is doing a lot of work, so it is worth unpacking what Wonderful says is in the box. The platform bundles four product lines that can be bought separately or combined: managed workflows that automate end-to-end business processes, productivity agents aimed at employees and decision support, AI-native applications intended to complement or replace legacy software outright, and conversational agents for customer-facing work. Underneath sits the shared layer that gives the product its name - enterprise context, integrations, security, and governed execution, applied consistently across everything built on top. Two architectural decisions matter more than the product taxonomy. First, the platform is model-agnostic. Wonderful's AI Gateway routes each request to whichever model suits the task, sending complex prompts to frontier models and cheap ones to smaller models, with per-team rules governing which groups can use which models and monitoring that flags cost spikes and unusual access patterns. Second, it is deployment-agnostic, running on any cloud environment including on-premise - a requirement, not a nicety, for regulated buyers in finance and healthcare. The developer-facing tooling is more conventional than the marketing suggests: team-scoped project folders, version control with rollback, and A/B testing so customers can compare agent variants before promoting one to production. That is deliberate. The company's argument is that enterprises already know how to run software; what they lack is a governed place to put AI. The thesis, in the founder's words. CEO and co-founder Bar Winkler frames the opportunity as a repeat of the cloud transition - with a specific warning attached about what happens if enterprises skip the platform layer. " Just as cloud platforms became the foundation of the modern enterprise, AI operating systems will become the foundation of every enterprise. Its customers are already proving that once AI reaches production in one part of the business, it quickly expands across the enterprise. Without a shared operating system, AI risks recreating the sprawl of traditional SaaS. - Bar Winkler, CEO and Co-founder, Wonderful Where Wonderful sits in a crowded, well-funded field. Wonderful is competing for the same enterprise budget as several companies that raised earlier and, in some cases, larger. The distinguishing variable is scope: most peers own one layer of the stack, while Wonderful is claiming all of them. Note that valuations below are private and reported, and ARR figures are company-disclosed rather than audited. | Company | Reported valuation | Latest round | Primary claim | | Wonderful | $5B (Sept 2026) | $550M Series C | Full-stack enterprise AI OS; agents, workflows, apps, governance | | Sierra | $15B+ (May 2026) | $950M, led by GV and Tiger Global | Customer-facing agents; outcome-based pricing, ~$200M ARR reported | | Glean | $7.2B (Dec 2025) | $150M Series F | Enterprise search and knowledge; $300M ARR reported May 2026 | | Decagon | $4.5B (Jan 2026) | $250M Series D | AI customer support automation; per-conversation pricing | The real product May be the engineers. The least software-like part of Wonderful's model is arguably the most important one. The company deploys forward-deployed engineering pods - teams that sit inside the customer's organization, connect the platform to real systems, and drive a first use case into production. Then, per the company's own description, they transfer the capability so the enterprise can build and operate on the platform independently. Insight Partners describes the compounding logic explicitly: reusable integrations and accumulated enterprise context are supposed to make each deployment cheaper than the last. This is the Palantir playbook, and in 2026 it has become close to standard among enterprise AI vendors. The reason is uncomfortable but well-documented. Gartner has forecast that more than 40% of agentic AI projects will be cancelled by the end of 2027, citing unclear business value, escalating costs, and inadequate risk controls. Multiple 2026 surveys put the share of enterprises that have genuinely scaled agents across the organization at roughly a quarter, even as the share running some kind of pilot approaches universality. The gap between a working demo and a governed production system is where enterprise AI budgets go to die. High-touch services close that gap. They also compress gross margins and make growth a function of headcount, which is precisely why Wonderful says a large share of this round goes toward expanding international FDE teams. Investors are underwriting a services-heavy business at a software multiple - a bet that the accumulated context and reusable integrations eventually flip the ratio. One more detail worth sitting with: Salesforce is now an investor. Salesforce sells its own agent platform into the same buyer. A strategic check from a competitor is usually a distribution signal, a hedge, or an early look at an acquisition target. Sometimes all three. What the round does not tell you. Wonderful has disclosed valuation, headcount, market count, and total capital raised. It has not disclosed annual recurring revenue, customer count, retention, or the software-versus-services split of its revenue - all of which peers at similar valuations have made public. Reported customer traction is described qualitatively ("hundreds of agents, systems, and workflows across a dozen verticals") rather than in contracted terms. Treat the $5 billion figure as a statement of investor conviction about the category, not as a proxy for realized revenue. Questions to ask before buying an "AI operating system" * What exactly happens to the agents and integrations you build if you leave the platform - do you retain the artifacts, or just the outputs? * Which models can you route to today, and how quickly are new frontier models added to the gateway? * How is the forward-deployed engagement priced, and at what point does it end rather than become a permanent line item? * Can the platform run in your existing security perimeter - including fully on-premise - without a feature downgrade? * What does governed execution actually enforce: approval gates, audit logs, data residency, per-team model permissions, or all four? * How does the vendor measure whether a deployed workflow is working, and will they contract to that metric? * Given that a large share of agentic projects are forecast to be cancelled, what is your own kill criterion and review date before you sign? Frequently asked questions. #Wonderful #enterprise AI #AI agents #Series C funding #AI operating system #Insight Partners #agentic AI #Salesforce