Denis Ignatovich, Co-founder and Co-CEO of Imandra, has over a decade of expertise in buying and selling, threat administration, quantitative modeling, and sophisticated buying and selling system design. Earlier than founding Imandra, he led the central threat buying and selling desk at Deutsche Financial institution London, the place he acknowledged the essential position AI can play within the monetary sector. His insights throughout this time helped form Imandra’s suite of monetary merchandise. Denis’ contributions to computational logic for monetary buying and selling platforms embrace a number of patents. He holds an MSc in Finance from the London Faculty of Economics and levels in Laptop Science and Finance from UT Austin.
Imandra is an AI-powered reasoning engine that makes use of neurosymbolic AI to automate the verification and optimization of complicated algorithms, notably in monetary buying and selling and software program techniques. By combining symbolic reasoning with machine studying, it enhances security, compliance, and effectivity, serving to establishments scale back threat and enhance transparency in AI-driven decision-making.
What impressed you and Dr. Grant Passmore to co-found Imandra, and the way did your backgrounds affect the imaginative and prescient for the corporate?
After school I went into quantitative buying and selling and ended up in London. Grant did his PhD in Edinburgh after which moved to Cambridge to work on functions of automated logical reasoning for evaluation of security of autopilot techniques (complicated algorithms which contain nonlinear computation). In my work, I additionally handled complicated algorithms with a number of nonlinear computation and we realized that there’s a deep connection between these two fields. The best way that finance was creating such algorithms was actually problematic (as highlighted by many information tales coping with “algo glitches”), so we got down to change that by empowering engineers in finance with automated logical instruments to convey rigorous scientific strategies to the software program design and improvement. Nevertheless, what we ended up creating is industry-agnostic.
Are you able to clarify what neurosymbolic AI is and the way it differs from conventional AI approaches?
The sphere of AI has (very roughly!) two areas: statistical (which incorporates LLMs) and symbolic (aka automated reasoning). Statistical AI is unimaginable at figuring out patterns and doing translation utilizing data it discovered from the info it was skilled on. However, it’s unhealthy at logical reasoning. The symbolic AI is nearly the precise reverse – it forces you to be very exact (mathematically) with what you’re attempting to do, however it will possibly use logic to cause in a means that’s (1) logically constant and (2) doesn’t require knowledge for coaching. The strategies combining these two areas of AI are known as “neurosymbolic”. One well-known utility of this strategy is the AlphaFold venture from DeepMind which just lately received the Nobel prize.
What do you assume units Imandra aside in main the neurosymbolic AI revolution?
There are numerous unimaginable symbolic reasoners on the market (most in academia) that concentrate on particular niches (e.g. protein folding), however Imandra empowers builders to investigate algorithms with unprecedented automation which has a lot higher functions and higher goal audiences than these instruments.
How does Imandra’s automated reasoning get rid of frequent AI challenges, corresponding to hallucinations, and enhance belief in AI techniques?
With our strategy, LLMs are used to translate people’ requests into formal logic which is then analyzed by the reasoning engine with full logical audit path. Whereas translation errors could happen when utilizing the LLM, the consumer is supplied with a logical clarification of how the inputs had been translated and the logical audits could also be verified by third get together open supply software program. Our final purpose is to convey actionable transparency, the place the AI techniques can clarify their reasoning in a means that’s independently logically verifiable.
Imandra is utilized by Goldman Sachs and DARPA, amongst others. Are you able to share a real-world instance of how your expertise solved a fancy downside?
An incredible public instance of the actual world affect of Imandra is highlighted in our UBS Way forward for Finance competitors 1st place win (the small print with Imandra code is on our web site). Whereas making a case research for UBS that encoded a regulatory doc that they submitted to the SEC, Imandra recognized a basic and delicate flaw within the algorithm description. The flaw stemmed from delicate logical circumstances that should be met to rank orders inside an order guide – one thing that might be inconceivable for people to detect “by hand”. The financial institution awarded us 1st place (out of greater than 620 firms globally).
How has your expertise at Deutsche Financial institution formed Imandra’s functions in monetary techniques, and what’s probably the most impactful use case you’ve got seen up to now?
At Deutsche Financial institution we handled a number of very complicated code that made automated buying and selling choices based mostly on varied ML inputs, threat indicators, and many others. As any financial institution, we additionally needed to abide by quite a few laws. What Grant and I spotted was that this, on a mathematical stage, was similar to the analysis he was doing for autopilot security.
Past finance, which industries do you see as having the best potential to learn from neurosymbolic AI?
We’ve seen AlphaFold get the Nobel prize, so let’s positively depend that one… In the end, most functions of AI will enormously profit by use of symbolic strategies, however particularly, we’re engaged on the next brokers that we are going to launch quickly: code evaluation (translating supply code into mathematical fashions), creating rigorous fashions from English-prose specs, reasoning about SysML fashions (language used to explain techniques in safety-critical industries) and enterprise course of automation.
Imandra’s area decomposition is a novel function. Are you able to clarify the way it works and its significance in fixing complicated issues?
A query that each engineer thinks about when writing software program is “what the sting circumstances?”. When their job is QA and they should write unit check circumstances or they’re writing code and interested by whether or not they’ve appropriately carried out the necessities. Imandra brings scientific rigor to reply this query – it treats the code as a mathematical mannequin and symbolically analyzes all of its edge circumstances (whereas producing a proof in regards to the completeness of protection). This function is predicated on a mathematical approach known as ‘Cylindrical Algebraic Decomposition’, which we’ve “lifted” to algorithms at massive. It has saved numerous hours for our prospects in finance and uncovered essential errors. Now we’re bringing this function to engineers in every single place.
How does Imandra combine with massive language fashions, and what new capabilities does this unlock for generative AI?
LLMs and Imandra work collectively to formalize human enter (whether or not it’s supply code, English prose, and many others), cause about it after which return the output in a means that’s simple to grasp. We use agentic frameworks (e.g. Langgraph) to orchestrate this work and ship the expertise as an agent that our prospects can use straight, or combine into their functions or brokers. This symbiotic workflow addresses lots of the challenges of utilizing LLM-only AI instruments and extends their utility past beforehand seen coaching knowledge.
What’s your long-term imaginative and prescient for Imandra, and the way do you see it remodeling AI functions throughout industries?
We expect neurosymbolic strategies would be the basis that paves the way in which for us to comprehend the promise of AI. Symbolic strategies are the lacking ingredient for a lot of the industrial functions of AI and we’re excited to be on the forefront of this subsequent chapter of AI.
Thanks for the good interview, readers who want to study extra ought to go to Imandra.
