CTGT wins Greatest Presentation Fashion award at VB Rework 2025

San Francisco-based CTGT, a startup centered on making AI extra reliable by means of feature-level mannequin customization, gained the Greatest Presentation Fashion award at VB Rework 2025 in San Francisco. Based by 23-year-old Cyril Gorlla, the corporate showcased how its know-how helps enterprises overcome AI belief boundaries by instantly modifying mannequin options as an alternative of utilizing conventional fine-tuning or immediate engineering strategies.

Throughout his presentation, Gorlla highlighted the “AI Doom Loop” confronted by many enterprises: 54% of companies cite AI as their highest tech threat in keeping with Deloitte, whereas McKinsey reviews 44% of organizations have skilled adverse penalties from AI implementation.

“A big a part of this convention has been in regards to the AI doom loop” Gorlla defined throughout his presentation. “Sadly, numerous these [AI investments] don’t pan out. J&J simply canceled a whole bunch of AI pilots as a result of they didn’t actually ship ROI attributable to no basic belief in these methods.”

Breaking the AI compute wall

CTGT’s strategy represents a major departure from typical AI customization strategies. The corporate was based on analysis Gorlla performed whereas holding an endowed chair on the College of California San Diego.

In 2023, Gorlla printed a paper on the Worldwide Convention on Studying Representations (ICLR) describing a technique for evaluating and coaching AI fashions that was as much as 500 occasions quicker than current approaches whereas attaining “three nines” (99.9%) of accuracy.

Fairly than counting on brute-force scaling or conventional deep studying strategies, CTGT has developed what it calls an “fully new AI stack” that basically reimagines how neural networks be taught. The corporate’s innovation focuses on understanding and intervening on the characteristic stage of AI fashions.

The corporate’s strategy differs basically from normal interpretability options that depend on secondary AI methods for monitoring. As an alternative, CTGT provides mathematically verifiable interpretability capabilities that eradicate the necessity for supplemental fashions, considerably decreasing computational necessities within the course of.

The know-how works by figuring out particular latent variables (neurons or instructions within the characteristic house) that drive behaviors like censorship or hallucinations, then dynamically modifying these variables at inference time with out altering the mannequin’s weights. This strategy permits corporations to customise mannequin conduct on the fly with out taking methods offline for retraining.

Actual-world purposes

Throughout his Rework presentation, Gorlla demonstrated two enterprise purposes already deployed at a Fortune 20 monetary establishment:

An e mail compliance workflow that trains fashions to know company-specific acceptable content material, permitting analysts to verify their emails towards compliance requirements in real-time. The system highlights doubtlessly problematic content material and gives particular explanations.

A model alignment software that helps entrepreneurs develop copy per model values. The system can recommend personalised recommendation on why sure phrases work nicely for a particular model and the best way to enhance content material that doesn’t align.

“If an organization has 900 use circumstances, they not need to fine-tune 900 fashions,” Gorlla defined. “We’re model-agnostic, to allow them to simply plug us in.”

An actual-world instance of CTGT’s know-how in motion was its work with DeepSeek fashions, the place it efficiently recognized and modified the options chargeable for censorship behaviors. By isolating and adjusting these particular activation patterns, CTGT was capable of obtain a 100% response fee on delicate queries with out degrading the mannequin’s efficiency on impartial duties like reasoning, arithmetic and coding.

Photographs: CTGT presentation at VB Rework 2025

Demonstrated ROI

CTGT’s know-how seems to be delivering measurable outcomes. Throughout the Q&A session, Gorlla famous that within the first week of deployment with “one of many main AI-powered insurers, we saved $5 million of legal responsibility from them.”

One other early buyer, Ebrada Monetary, has used CTGT to enhance the factual accuracy of customer support chatbots. “Beforehand, hallucinations and different errors in chatbot responses drove a excessive quantity of requests for reside help brokers as prospects sought to make clear responses,” stated Ley Ebrada, Founder and Tax Strategist. “CTGT has helped enhance chatbot accuracy tremendously, eliminating most of these agent requests.”

In one other case research, CTGT labored with an unnamed Fortune 10 firm to reinforce on-device AI capabilities in computationally constrained environments. The corporate additionally helped a number one laptop imaginative and prescient agency obtain 10x quicker mannequin efficiency whereas sustaining comparable accuracy.

The corporate claims its know-how can cut back hallucinations by 80-90% and allow AI deployments with 99.9% reliability, a important issue for enterprises in regulated industries like healthcare and finance.

From Hyderabad to Silicon Valley

Gorlla’s journey is itself exceptional. Born in Hyderabad, India, he mastered coding at age 11 and was disassembling laptops in highschool to squeeze out extra efficiency for coaching AI fashions. He got here to america to check on the College of California, San Diego, the place he obtained the Endowed Chair’s Fellowship.

His analysis there centered on understanding the elemental mechanisms of how neural networks be taught, which led to his ICLR paper and finally CTGT. In late 2024, Gorlla and co-founder Trevor Tuttle, an knowledgeable in hyperscalable ML methods, have been chosen for Y Combinator’s Fall 2024 batch.

The startup has attracted notable traders past its institutional backers, together with Mark Cuban and different outstanding know-how leaders drawn to its imaginative and prescient of constructing AI extra environment friendly and reliable.

Funding and future

Based in mid-2024 by Gorlla and Tuttle, CTGT raised $7.2 million in February 2025 in an oversubscribed seed spherical led by Gradient, Google’s early-stage AI fund. Different traders embody Normal Catalyst, Y Combinator, Liquid 2, Deepwater, and notable angels reminiscent of François Chollet (creator of Keras), Michael Seibel (Y Combinator, co-founder of Twitch), and Paul Graham (Y Combinator).

“CTGT’s launch is well timed because the business struggles with the best way to scale AI throughout the present confines of computing limits,” stated Darian Shirazi, Managing Companion at Gradient. “CTGT removes these limits, enabling corporations to quickly scale their AI deployments and run superior AI fashions on units like smartphones. This know-how is important to the success of high-stakes AI deployments at massive enterprises.”

With AI mannequin dimension outpacing Moore’s Regulation and advances in AI coaching chips, CTGT goals to concentrate on a extra foundational understanding of AI that may deal with each inefficiency and more and more complicated mannequin choices. The corporate plans to make use of its seed funding to broaden its engineering workforce and refine its platform.

Every finalist offered to an viewers of 600 business decision-makers and obtained suggestions from a panel of enterprise capital judges from Salesforce Ventures, Menlo Ventures, and Amex Ventures.

Learn in regards to the different winners Catio and Solo.io. The opposite finalists have been Kumo, Superduper.io, Sutro and Qdrant.

Editor’s be aware: As a thank-you to our readers, we’ve opened up early fowl registration for VB Rework 2026 — simply $200. That is the place AI ambition meets operational actuality, and also you’re going to wish to be within the room. Reserve your spot now.