Blitz co-founders on building a company from the data that kept getting them banned at sportsbooks

Devon Sinha and Tejas Srinivasan met freshman year at Duke as neighbors only two rooms apart, and spent the next several years running up edges that sportsbooks couldn't see.

It started during senior year with PrizePicks. Sinha built a theory around extreme correlative scenarios operators weren't pricing correctly: what happens in baseball when a team scores zero runs in the first inning? He deposited $100, placed a parlay, and hit for $25,000. He ran the same logic across every game for a month. That $100 became $70,000—and then he got limited.

After graduating, both landed software engineering jobs in Seattle—Sinha at AWS, Srinivasan at Microsoft. Washington State has no online sports betting, so they went in person. Each morning, they'd take turns driving to Muckleshoot Casino (a Caesars affiliate) to place their picks for the full day's slate. The strategy was to build their own historical databases, find scenarios books couldn't price, and load up on volume. One play involving unders on RBIs got large enough that Caesars banned them—and, Sinha said, later removed that market category from the site entirely.

Eventually the math changed. "Look, I'm not gonna have time for this," Sinha said. "I can't just be sitting around trying to snipe lines." But the infrastructure they'd built up of historical databases, automated validation, multi-source cross-referencing was still there. That became Blitz.

At a high level, Blitz is what the co-founders call a "sports intelligence infrastructure," or in other words, an AI-powered platform built on what they describe as the most complete and cleanest historical sports dataset available, covering eight leagues from inception. On top of that: a conversational chatbot for research, a real-time insights engine that surfaces in-game storylines, and auto-generated content for previews, recaps, and summaries.

The early consumer version, which was initially rejected by the App Store over betting-related content, was deliberately shelved. The market that stuck was B2B—broadcasters, analysts, and content creators who'd been cobbling together research from clunky dashboards. "It's all these people who are using tools like Savant and FanGraphs and a bunch of other MLB dashboards that were clunky," Srinivasan said, "now coming to Blitz to ask a very simple question in English and just get back the data and results really quickly." Ryan Spilborghs, a former Colorado Rockies outfielder now broadcasting on Apple TV, adopted the product for his own pre-game research and joined as an advisor.

The data itself is both the hardest part and the moat. Official providers like SportRadar prioritize live feeds over historical accuracy, Devon says. Blitz found data from a major provider listing a player with 135 minutes in a game. Their response was to pull from multiple sources, write automated tests to cross-validate every entry, and keep building until head-to-head comparisons showed league-built chatbots failing basic statistical queries 80–90% of the time. "You can't really take any shortcuts in this space," Srinivasan said.

Devon phrased the AI architecture as  "deterministic stacks, generative voice." Historical queries always route to Blitz's own validated database. Live data (injuries, current betting lines) gets fetched fresh. Getting that routing right, he argued, is the kind of sports-specific nuance you can't solve by attaching an LLM to a data feed.

Blitz bootstrapped on betting winnings through most of 2025, then raised a friends-and-family round in early 2026. A venture round is expected in the coming months as the company expands league coverage and grows the engineering team.

Listen to the full podcast on YouTube, Spotify, and Apple Podcasts.