As software systems become more complex and AI tools generate code faster than ever, a fundamental problem is becoming worse: Engineers drowning in debugging workSpend half their time discovering the causes of software failures instead of building new products. The challenge has become so severe that it is creating a new class of tooling – AI agents that can diagnose production failures in minutes instead of hours.
Deductive AIA startup that emerged from stealth mode on Wednesday believes it has found a solution by applying reinforcement learning — the same technology that powers game-playing AI systems — to the messy, high-stakes world of production software events. The company announced that it has raised $7.5 million in seed funding CRVwith the participation of Databricks Ventures, Thomvest VenturesAnd primesetcommercialize what is called "AI SRE Agent" Which can help diagnose and fix software failures at machine speed.
The pitch resonates with the growing frustration inside engineering organizations: Modern observation tools can show that something broke, but they rarely explain why. Even when a production system fails at 3 a.m., engineers must face hours of manual detective work, cross-referencing logs, metrics, deployment history, and code changes across dozens of interconnected services to identify the root cause.
"The complexities and inter-dependencies of modern infrastructure mean that investigating the root cause of an outage or incident can feel like finding a needle in a haystack, except that haystack is the size of a football field, it’s made up of millions of other needles, it’s constantly shuffling, and setting fire to itself – and every second you can’t find it equates to lost revenue." Sameer Aggarwal, co-founder and chief technology officer of Deductive, said in an exclusive interview with VentureBeat.
Deductive’s system creates what the company calls "knowledge graph" Which maps relationships across codebases, telemetry data, engineering discussions, and internal documentation. When an incident occurs, multiple AI agents work together to generate hypotheses, test them against live system evidence, and converge on the root cause – mimicking the investigation workflow of experienced site reliability engineers, but completing the process in minutes instead of hours.
The technology has already shown measurable impact on some of the world’s most demanding production environments. DoorDash’s advertising platformwhich runs real-time auctions that must complete in less than 100 milliseconds, has integrated Deductive into its incident response workflow. The company has set an ambitious 2026 goal of resolving production incidents within 10 minutes.
"Our ad platform is operating at a speed where manual, slow-moving testing is no longer viable. Every minute of downtime directly impacts company revenue," Shahrooz Ansari, senior director of engineering at DoorDash, said in an interview with VentureBeat. "Deductive has become a vital extension of our team, rapidly synthesizing signals across dozens of services and uncovering critical insights in minutes."
Doordash Deductive is estimated to have given rise to approximately 100 production incidents over the past few months, meaning over 1,000 hours of annual engineering productivity and revenue impact "in millions of dollars," According to Ansari. intelligence company on location truthfullyDeductive reduced the time to diagnose Apache Spark job failures by 90% – transforming a process that previously took hours or days to one that is completed in less than 10 minutes – while generating more than $275,000 in annual savings.
Why is AI-generated code causing a debugging crisis?
The timing of Deductive’s launch reflects a growing tension in software development: AI coding assistants are enabling engineers to generate code faster than ever, but the resulting software is often harder to understand and maintain.
"vibe coding," A term popularized by an AI researcher Andrej KarpathyAI refers to using natural-language prompts to generate code through assistants. While these tools speed up development, they can introduce what Agrawal points out "Redundancy, breaking architectural limitations, assumptions, or neglected design patterns" Which accumulates over time.
"Most AI-generated code still introduces redundancies, breaks architectural boundaries, makes assumptions, or ignores established design patterns," Aggarwal told VentureBeat. "In many ways, we now need AI to help us clean up the mess it is creating."
The claim that engineers spend almost half their time debugging is not an exaggeration. The Computing Machinery Association reports that developers spend 35% to 50% of their time is spent validating and debugging softwaremore recently, Harness’ Software Delivery Status 2025 The report found that 67% of developers are spending more time debugging AI-generated code.
"We have seen world class engineers spending half their time debugging instead of building," said Rakesh Kothari, co-founder and CEO of Deductive. "And as vibe coding generates new code at a rate we’ve never seen before, this problem is only going to get worse."
How Deductive’s AI agents actually investigate production failures
Deductive’s technical approach is quite different from the AI features being added to existing observability platforms datadog Or new relicMost of those systems use large language models to summarize data or identify correlations, but they don’t have what Agarwal says "code-aware logic"- The ability to understand not only that something broke, but also the ability to understand why the code behaves that way.
"Most enterprises use multiple observability tools across different teams and services, so no single vendor has a single holistic view of how their systems behave, fail, and recover – nor are they able to combine this with an understanding of the code that defines system behavior," Aggarwal explained. "These are key elements to resolving software incidents and it really fills the gap."
The system connects to existing infrastructure using read-only API access to observability platforms, code repositories, incident management tools, and chat systems. It then continuously builds and updates its knowledge graph, mapping dependencies between services and tracking deployment history.
When an alert is activated, Deductive launches what the company describes as a multi-agent investigation. Different agents specialize in different aspects of the problem: one may analyze recent code changes, another examines trace data, while a third correlates the time of the event with recent deployments. Agents share findings and iteratively refine their hypotheses.
The key difference from rule-based automation is the use of reinforcement learning by deductive. The system learns from each incident which investigation steps led to the correct diagnosis and which steps were incomplete. When engineers provide feedback, the system incorporates that signal into its learning model.
"Every time he observes an investigation, he knows which steps, data sources, and decisions led to the correct results," Aggarwal said. "It’s learning how to think about problems, not just point to them."
At DoorDash, the recent latency increase in the API initially appeared to be an isolated service issue. Deductive investigation revealed that the root cause was actually timeout errors from a downstream machine learning platform undergoing deployment. The system connected these dots by analyzing log volumes, traces, and deployment metadata across multiple services.
"Without Deductive, our team would have to manually correlate latency spikes across all logs, traces, and deployment history," Ansari said. "Deductive was able to explain not only what changed, but how and why it affected production behavior."
The company keeps humans in the loop for now
While Deductive’s technology could theoretically directly improve production systems, the company has deliberately opted to keep humans in the loop – at least for now.
"While our system is capable of deep automation and can improve production, currently, we recommend precise improvements and mitigations that engineers can review, verify, and implement." Aggarwal said. "We believe maintaining a human in the loop is essential for trust, transparency, and operational security."
However, he acknowledged that "Over time, we think deeper automation will come and how humans work in the loop will evolve."
Databricks and ThoughtSpot giants bet on logic rather than observability
The founding team brings deep expertise from building some of Silicon Valley’s most successful data infrastructure platforms. Aggarwal received his Ph.D. Earned. at UC Berkeley, where he composed blinkdbAn efficient system for approximate query processing. He was among the first engineers databrickswhere he helped build apache sparkKothari was an early engineer thoughtspotWhere he led teams focused on distributed query processing and large-scale system optimization.
The investor syndicate reflects both technical credibility and market opportunity. Beyond CRV max gazorThe round included participation from ion stoicaFounder of Databricks and Anyscale; Ajit SinghFounder of Nutanix and ThoughtSpot; And Ben SigelmanFounder of Lightstep.
Instead of competing with platforms like datadog Or pagerdutyDeductive positions itself as a complementary layer that sits on top of existing tools. The pricing model reflects this: instead of charging based on data volume, a deductive fee is charged based on the number of incidents examined, as well as a base platform fee.
The company offers both cloud-hosted and self-hosted deployment options and emphasizes that it does not store customer data on its servers or use it to train models for other customers – an important assurance given the proprietary nature of both the code and production system behavior.
With new capital and early customer attraction in companies like Doordash, truthfullyAnd Kumo AIDeductive plans to expand its team and deepen the system’s reasoning capabilities, from reactive incident analysis to proactive prevention. Near-term perspective: Helping teams anticipate problems before they happen.
DoorDash’s Ansari offers insightful support for the state of the technology today: "Investigations that were previously manual and time-consuming are now automated, allowing engineers to shift their energies toward prevention, business impact, and innovation."
In an industry where every second of downtime translates into lost revenue, the shift from firefighting to building construction looks less like a luxury and more like table stakes.