A DeepMind Spinout Wants to Turn Scientific Papers Into Queryable Tables

The Montreal startup, founded by former DeepMind researchers, automates the grueling data extraction at the heart of drug discovery. Cohere bought the company in 2026.

About Reliant AI

Published

In the quiet, expensive rooms where new drugs are conceived, the first step is often the most tedious. Teams of analysts, interns, and graduate researchers spend weeks, sometimes months, systematically reviewing thousands of scientific papers and regulatory documents. They are hunting for evidence: a trial result, a mechanism of action, a safety signal. It is work measured in coffee cups and highlighters, a bottleneck of human attention that Karl Moritz Hermann decided was ripe for an algorithmic takeover.

Hermann, a former senior research scientist at Google DeepMind, co-founded Reliant AI in 2023 to tackle that bottleneck. The Montreal-based startup’s first product, Reliant Tabular, is an AI workbench designed to let life-science analysts query vast collections of documents and extract structured information, even when the source material is inconsistently labeled or purely unstructured [Reliant AI, August 2024]. In August 2024, the company emerged from stealth with an $11.3 million seed round, part of a total $13.5 million raise, to prove that the most valuable data in biopharma is often trapped in PDFs [TechCrunch, August 2024] [BetaKit, August 2024].

The Wedge: Automating the Systematic Review

Reliant’s initial wedge is the systematic literature review, a foundational but brutally manual process in drug development. The company claims its technology combines the Llama 3.1 large language model with proprietary techniques for document extraction and analysis, moving beyond simple keyword search to understand context and relationships within scientific text [TechCrunch, August 2024]. The output is not just a list of relevant papers, but structured tables and what the company calls “densified Knowledge Graphs” that map connections between trials, drug mechanisms, and outcomes [Reliant AI, retrieved 2025].

The early value proposition is straightforward: time saved. In one cited example, a client reduced the time spent per document from approximately 120 minutes to 20 minutes using Reliant Tabular, processing 51 studies through the platform [TipRanks.com, retrieved 2026]. Another user reported saving a full week of work on a project preparing materials for primary market research [Reliant AI, retrieved 2026]. For an industry where speeding up R&D by a single month can be worth millions, that’s a compelling opening argument.

A Team Built for the Technical Grind

The founding team reads like a blueprint for this specific problem. Hermann and co-founder Marc Bellemare, the chief scientific officer, are both former DeepMind researchers; Bellemare also holds a Canada CIFAR AI Chair at the Mila institute in Montreal [Reliant AI, August 2024]. The third co-founder, Richard Schlegel, head of commercial, came from EY-Parthenon, providing the commercial lens to ground the technical ambition [Reliant AI, August 2024]. This blend of deep AI research and practical business strategy was convincing enough to attract a seed round co-led by Tola Capital and Inovia Capital, with participation from angels including former Datadog executive Amit Agarwal and AI pioneer Yoshua Bengio [BetaKit, August 2024].

Founder Role Key Background
Karl Moritz Hermann Co-founder & CEO Former Senior Research Scientist, Google DeepMind [Reliant AI, August 2024]
Marc Bellemare Co-founder & Chief Scientific Officer Former DeepMind Researcher; Canada CIFAR AI Chair at Mila [Reliant AI, August 2024]
Richard Schlegel Co-founder & Head of Commercial Former EY-Parthenon [Reliant AI, August 2024]

From Stealth Startup to Cohere Acquisition

The company’s trajectory accelerated quickly. By the time of its reported acquisition by AI company Cohere in May 2026, Reliant AI had grown to 41 employees and was working with global pharmaceutical companies and leading biotechs, including named customers GSK, Medicus Pharma, and Kyowa Kirin [PitchBook, retrieved 2026] [Analytica World, November 2026] [Baker McKenzie, May 2026]. The acquisition was framed as a vertical expansion for Cohere, integrating Reliant’s team, specialized biomedical datasets, and domain-optimized technology into a new offering dubbed “North for Pharma” [LinkedIn, retrieved 2026] [The Healthcare Technology Report, retrieved 2026].

Hermann and Bellemare joined Cohere as Vice Presidents of AI Verticalizations and Modeling, respectively, suggesting the technology and approach would live on within a larger platform [Morningstar, May 2026]. The deal underscores a strategic truth: highly specialized AI for regulated, data-rich industries like biopharma can become valuable acquisition targets even before reaching massive independent scale.

The Incumbent It Must Beat

The most credible competitive pressure for a tool like Reliant Tabular isn’t another AI startup, but the entrenched incumbent: the manual workflow itself, often supported by a patchwork of generic search databases and an army of consultants. The unit economics of that incumbent are built on billable hours and low-cost labor, not software margins.

A back-of-the-envelope calculation illustrates the stakes. If a mid-sized biotech employs five analysts who each spend a quarter of their time on systematic reviews, that’s roughly one full-time equivalent, costing at least $150,000 annually in salary and overhead. Reliant’s software, priced as enterprise SaaS, needs to cost meaningfully less than that while delivering the same or better output faster. The 80% reduction in time per document they cite suggests it can, but the real test is whether it can consistently do so across thousands of documents and complex, novel queries where the AI might hallucinate or miss nuance.

To win, Reliant’s technology, now under Cohere’s banner, must prove it’s more reliable and comprehensive than a diligent human with a highlighter and a spreadsheet. It must become the trusted system of record for evidence, not just a faster first pass. That’s a higher bar, but one that justifies the acquisition price and the seed investors’ faith. For an industry betting billions on molecules, trusting an algorithm with the foundational literature might be the biggest leap of all.

Sources

  1. [Reliant AI, August 2024] Seed Announcement | https://www.reliant.ai/seed-announcement
  2. [TechCrunch, August 2024] Reliant’s paper-scouring AI takes on science’s data drudgery | https://techcrunch.com/2024/08/20/reliant-ai/
  3. [BetaKit, August 2024] Ex-Google DeepMind leaders bring Reliant AI out of stealth with $15.4-million CAD seed round | https://betakit.com/ex-google-deepmind-leaders-bring-reliant-ai-out-of-stealth-with-15-4-million-cad-seed-round/
  4. [TipRanks.com, retrieved 2026] Reliant AI client case study | https://www.tipranks.com
  5. [Reliant AI, retrieved 2025] Complex research made simple | https://www.reliant.ai/
  6. [PitchBook, retrieved 2026] Reliant AI employee count | https://pitchbook.com
  7. [Analytica World, November 2026] Cohere acquires biopharma AI specialist Reliant AI | https://www.analytica-world.com/en/news/1188778/cohere-acquires-biopharma-ai-specialist-reliant-ai.html
  8. [Baker McKenzie, May 2026] Cohere acquisition announcement | https://www.bakermckenzie.com
  9. [LinkedIn, retrieved 2026] Tola Capital post on acquisition | https://www.linkedin.com
  10. [The Healthcare Technology Report, retrieved 2026] Coverage of Cohere-Reliant AI deal | https://thehealthcaretechnologyreport.com
  11. [Morningstar, May 2026] Reliant AI founders join Cohere | https://www.morningstar.com

Read on Startuply.vc