Bin Ni's DeepMorph reports $8.575M sold toward a $9M target

Why it matters

Ni is betting that enterprise AI coding will create a separate market for software that verifies business behavior. The filing reports $8.575M sold and $425,000 remaining in the $9M offering, but identifies neither the four investors nor DeepMorph's valuation.

DeepMorph, the enterprise software startup led by former Google code-AI executive Bin Ni, has sold $8.575 million of a targeted $9 million financing, according to a Form D filed September 9th. The Newark, California-based company is building a validation layer designed to check AI-generated software against business rules extracted from an enterprise's existing code.

Honeycomb identified Ni as Google's Head of AI Developer Assistant in its account of the HumanX conference in March 2025. He was also among the authors of Google's 2024 CodeGemma technical report, which covered open models built for code completion and generation.

DeepMorph takes Ni downstream from generating code into deciding whether generated code can be trusted. That distinction matters inside large enterprises, where a rewrite can pass conventional tests while changing pricing rules, approval logic or regulatory controls accumulated over decades.

A $9M target with $425,000 left

DeepMorph's offering began with a first sale on August 25th, roughly two weeks before the filing. The filing reports four participating investors and $8.575 million sold, equal to 95.3% of the $9 million offering. It lists $425,000 as remaining.

The securities are structured as a Simple Agreement for Future Equity, or SAFE, alongside rights to acquire securities. The filing classified the offering under Rule 506(b), which generally prohibits general solicitation or advertising. It also reported no sales commissions or finder's fees.

The four investors are not named. The filing also does not identify a lead investor or provide a valuation, SAFE conversion cap, discount or other economic terms that would show the price those investors placed on DeepMorph. The documented amount sold is therefore $8.575 million, with the offering still $425,000 short of its target when the notice was filed.

DeepMorph was incorporated in Delaware in 2025, and the filing lists Newark, California, as its principal business location. It identifies Ni as an executive officer and director; he signed as chief executive officer on September 8th. Ni is the sole person the filing lists in the executive officer, director or promoter categories.

The available filing materials do not disclose DeepMorph's revenue, ARR, customer count or headcount. The company website does not publish pricing.

DeepMorph wants to recover the rules before agents rewrite them

DeepMorph describes its software as "Validation AI." According to the company's product description, the platform reverse-engineers legacy repositories, databases and documentation to construct what it calls a deterministic record of an organization's business logic. Coding agents can then be checked against that record during planning, implementation, testing and production.

The approach addresses a specific enterprise problem. Legacy applications often contain rules that are poorly documented outside the code itself. An AI agent tasked with replacing or extending those applications may produce plausible software without preserving every exception and dependency encoded in the old system. DeepMorph is trying to turn those hidden rules into explicit constraints before a rewrite reaches production, while leaving human engineers responsible for final decisions.

Ni said in an August LinkedIn post introducing DeepMorph that he built the company with former Google and DeepMind research leads. He described validation and governance as constraints on enterprise software delivery as agents produce code faster. Neither the post nor the Form D identifies the other co-founders.

DeepMorph advertises enterprise design partnerships and a 48-hour, air-gapped pilot. A LinkedIn post about DeepMorph says an unnamed insurance provider reduced a flagship rewrite from months to weeks using the product; the customer, timeline and result have not been independently verified.

DeepMorph's website says its engine mapped more than 1.5 million lines of code in hours. That figure is a company marketing claim and has not been independently verified.

The design-partner model gives DeepMorph access to the large, complex codebases it needs to test whether extracted business logic is complete enough to govern an agent's work. Participating enterprises can shape the product around their migration requirements. The $8.575 million gives Ni capital to pursue those deployments, although the available materials offer no evidence yet of a repeatable sales process.

Code validation is drawing venture money

Several funded companies are pursuing adjacent parts of the same problem as AI tools generate a larger share of production code.

Qodo raised $70 million in March 2026 for AI-assisted code review, testing and governance. Greptile raised $25 million in September 2025 for repository-aware review of code written by people and AI systems.

Symbiotic Security raised $10 million in January 2026 to build security checks into AI code generation. Its workflow applies policies before generation, scans the resulting code and sends detected vulnerabilities through automated remediation.

RevEng.AI raised $15 million in May 2026 for a different checkpoint: analyzing compiled binaries to identify vulnerabilities, hidden components and suspicious behavior, including in software whose source code is unavailable.

DeepMorph is positioning itself around business-logic preservation. Security products examine whether generated code introduces vulnerabilities, while binary-analysis software inspects the finished program. DeepMorph's stated job is to determine whether a modified or replacement system retains the rules on which the business depends. Legacy-modernization vendors such as Renovix and Xamun also market AI-assisted analysis of older systems, leaving DeepMorph to prove that its validation layer can remain useful after an initial migration.

Ni's background gives him direct experience with systems that generate and assist with code. DeepMorph is his attempt to supply a control layer as those tools move from individual developer assistance into enterprise software production.

Extracting deterministic ground truth from repositories shaped by years of patches, undocumented exceptions and organizational turnover remains the difficult part. A missed rule can be as consequential as an invented one. DeepMorph's filing records four participating investors and $8.575 million sold, while their identities, the company's valuation and evidence of commercial traction remain private.