← Sticker Wall AGENT · DETAIL

Outsourced projects for formal verification of financial algorithms using Imandra at RMB 20,000 per project

Workflow: Every morning, screen verification requests for smart contracts or trading algorithms posted by quant teams and fintech

AGENT

Key Fields

FIELD STAMPS

IndustryFintech

RegionGlobal（海外）

ScaleSME

ChannelOnline

🔧 Workflow

Every morning, screen verification requests for smart contracts or trading algorithms posted by quant teams and fintech companies, use Imandra to build formal models and run automated reasoning, and break down algorithm behavior into a set of provable properties. In the afternoon, compile reports on counterexamples and boundary conditions, marking under which inputs each rule is violated. The input is the client's algorithm code and business rules; the output is a reproducible verification report and remediation suggestions. Human experts are responsible for the final judgment on conclusions and for explaining the business meaning of each counterexample to the client, while AI is responsible for exhaustively exploring the state space.

🛠 Setup Requirements

You need a foundation in functional programming and mathematical logic, be familiar with the OCaml or Python interfaces, and cannot substitute this with no-code tools; it is recommended to have studied discrete mathematics before starting. The learning path is to closely read Imandra's official documentation and example library for about two to four weeks, then reproduce official cases and write a retrospective note. On the tooling side, register an Imandra account and configure a cloud reasoning environment; there is no need to build your own hardware, and a regular local laptop is sufficient. The main investment is learning time and case accumulation, not equipment cost.

🧰 Toolchain

🔧 Imandra formal verification platform
🔧 GitHub code hosting and version management
🔧 Python or OCaml development environment
🔧 Notion for writing verification reports

💰 Revenue

① Service fees for financial algorithm formal verification projects (main income): quant teams and fintech companies pay per project, about RMB 20,000/project × 1-2 orders per month = RMB 20,000-40,000 in monthly income; 67% of Tier-1 banks have already pushed Agentic AI into production environments (media estimate, not independently reviewed), and the weight of this revenue stream in total income has not been broken out (self-reported by the merchant, not independently verified by a third party; time point is 2026); ② High ticket size for complex verification projects: complex projects charge more than RMB 50,000, but there is no data on how many orders were taken or what share of total revenue they account for; ③ Reuse of verification model templates: the delivery cycle for subsequent similar projects can be shortened by more than half, but the number of reusable orders and the pricing after reuse have not been disclosed, and the weight is also blank; ④ Opportunity item—annual compliance audit verification subscription: the financial AI Agent market is sized at $985.4M in 2026 (media estimate, not independently reviewed), charging by annual subscription or per-verification; this revenue has not yet been estimated.

💸 Cost

Imandra is paid according to its official subscription tiers; the individual tier is about several hundred RMB per month, and cloud computing power fluctuates with the volume of reasoning tasks. Including GitHub private repositories, collaboration tools, and report formatting tools, the monthly cost is about RMB 500 to 1,000, less than 5% of revenue.

⏱ Time Investment

During the learning period, 2 to 3 hours per day for about one month; after formally taking orders, each order requires about 15 to 25 hours to complete modeling, verification, and report writing, equivalent to about 3 hours per day. Reserve half a day each week to maintain the portfolio and respond to inquiries; this is a semi-project-based flexible work rhythm.

🚀 Getting Started

Step one: closely read Imandra's official website documentation and get the official financial case demo running, then organize the modeling process, counterexample results, and explanations item by item into a Chinese technical blog published publicly, forming searchable professional credentials. Then take small code audit orders on Upwork and in quant communities, exchanging below-market first-order pricing for real delivery records and client reviews, and within three months accumulate enough for three cases before returning to full-price orders.

🔑 Keys to Success

✅ Formal verification is a scarce skill with very few global competitors, hence high ticket sizes
✅ Building trust with a public demo portfolio and retrospective articles is more effective than mass-sending resumes
✅ Human experts judge and explain report conclusions to clients, while AI is responsible for exhaustively finding counterexamples; the division of labor is clear
✅ Reusing verification model templates and rule libraries can shorten the delivery cycle for subsequent similar projects by more than half

⚠️ 风险

⚠️ The learning curve is steep; those without a sufficient mathematical foundation may not win any orders for months
⚠️ Financial institutions have long verification cycles and strict acceptance, so payment collection may be delayed by several months
⚠️ Errors in delivered conclusions may affect client fund safety, so liability boundaries must be defined in contracts or professional liability insurance purchased

📌 Real Cases

📌 Imandra's official website discloses that its technology has been used for algorithm verification in finance and other critical fields, with the goal of using automated reasoning to ensure systems are safe, explainable, and fair, and delivering externally through a Reasoning as a Service model
📌 The third-party review platform opentools.ai noted in its 2026 review that Imandra provides neuro-symbolic automated reasoning, specifically serving verifiable, compliant, and resilient financial and safety-critical systems, reflecting that this direction already has a mature commercialization form
📌 At the 2026 China International Financial Exhibition, Alibaba Cloud's Dianjin agent announced the opening of more than 100 financial capabilities including wealth management, credit risk control, and insurance business development, and Sunline also launched a multi-agent collaborative financial agent product, indicating that financial institutions have sustained external procurement demand for verification and compliance services when pushing AI into core processes

Sources