Igor Ševo, Ph.D.

Biography, work and philosophy

Quick info

Dear client, student, friend or acquaintance, I am humbled by your interest in my work and person and so, in this regard, I provide you with links to specific information you may be looking for.

If you are interested in my software development CV, related technical competencies, education and work experience, for professional reasons, I recommend you skip directly to Software Development. For information about my artificial intelligence consulting services and enterprise AI implementation work, visit Consulting & AI.

An image of me where I appear well-dressed and pensive.

If you care more about my artistic endeavors, enjoy my music or want to know more about me as an artist and musician, then go to Music, writing and art. If it's the case that you're interested in martial-arts training, physical arts or any discussion of athletics, biomechanics or kinesiology, you can skip down to Physical arts. For information about my Brazilian Jiu-Jitsu club, visit Gavran BJJ.

Finally, if you've read some of my written work or listened to my philosophy lectures and are interested to know more about me, then read on through Philosophy, but do be ready for unbridled humor, gratuitous references, and unredacted personal ramblings of an aspirant polymath who resigned from the position of a university professor in pursuit of individual academic goals.

Philosophy in short – a soul's feature vector

What is written ahead is most probably a very poor attempt at abridging an entire philosophy into a few paragraphs of a deliberate satire standing for a personal introduction. You are not judged for proceeding with resolute skepticism and utmost intellectual reservation, but I still dare demand you be not dismayed by the awkward and incongruent language, for what I lack in humor and wit, I redeem with ostentatious intellectual narcissism.

However vain it might be of me to bring up a website merely to present my lowly person to the patently indifferent world, our souls have already inextricably become property of the public domain, so if we are mandated to compress, prune, quantize and dilute our essence, if we are to butcher our being and pack its infinite complexity into a tiny pre-designated manifold, I see no reason why not submit to futile hope and do it on our hypothetical own merit and of our theoretical own accord. Whether for an abundance of unsubstantiated intellectual arrogance or a deluded sense of digital freedom, I refuse, even if only in part so, the corporate owners of the virtual realm to scaffold for me the frames in which I should trap my humble me, so—behind a verbal façade extraordinaire—I introduce myself here. Although, I am perfectly and bleakly aware that whether you swallow swords or fervor fellatio, you still are working the same apparatus.

When language and thought have been distorted by the spreading contagion of epistemic relativism, manifesting intellectual sovereignty and pursuing truth become a senseless play of catering to audiences. This ostensible freedom bound by the critic’s mental vocabulary leaves me with the inexorable question of whom this introduction is intended for. Should I stick a dick joke mid the text or alliterate an academic apology to excuse such unwarranted toilet humor? It seems, if one exists in both worlds, he will ultimately belong to neither. Yet, this delicate balance of wisdom is performed right there on the edge, between the eternal child and the wise old man, where one can become everything without succumbing to naivety. It must take at least a modicum of intellectual audacity to step on that line, but not beyond, to navigate the vast dimensions of moral, logical, and physical space in perpetual pursuit for that one point exactly where all frontiers cross, from which to finally koan: “I am all, but I am me”. Though that spot is almost certainly endlessly out of reach, I imagine it is a lonely place, the last exit of the last cave, where no one has gone before. Every little bit of knowledge one absorbs brings him closer to knowing the universe, the entropy of the observed world declining towards predictable certainty, as yours soars in the wake; but it is a dreadful affair to want to know—it means to change, to die, in a way. Conscience doth make cowards of us all.

It may be an endless and futile pursuit, da capo al coda, ad infinitum, but it is one that nonetheless offers meaning; and even if the part repeats endlessly, and the cadence is forever past the coda, there is a new line, a new voice of the composer from beyond the paper, that can be heard mellowing mellifluously through the seemingly chaotic counterpoint of nature. For the pious or religious, I do not mean god—I mean whatever could exist beyond the confines of what we represent as mind, space and time, be it objectifiable or counterfactual, meta-conscious or metastatic—I am not entertaining theology, I am merely poeticizing metaphysics, l'art pour l'art, though poetry is often nothing more than a smart way to conceal a fart joke. Though it may seem otherwise, this unmistakably inept verbal egotism is not a reflection of my personality, only an unavoidable side-effect of attempting to layer a personal introduction beneath a guise of verbose cynical philosophy. If you can see in this sarcastic recital of clumsy arrogance a benign sort of idealist humor, and somewhere past it my humble facetious character, then I sincerely hope we meet in person. If you, however, believe me full of proverbially proverbial shit, make solace in the fact that for that very reason I think the same of you.

To divorce philosophy from practical life would be to commit crime against one’s own identity—admittedly or not, we all act out and through our principles, and so their consignation to the domain of afterthought is nothing more than a denial of one’s own being, a feeble proclamation of meekness before the feet of a peremptory social leviathan, whose callousness is built by exactly those avoidant inanities. So, no, I will not write a simple biography, devoid of me, but riddled with trivia, but rather assume that whoever is inclined to read my writing is both willing to know me and apt enough to hear the undertones, although I indulge occasionally the proposition of the statement “the little details are by far the most important”. Therefore, I dare stubbornly claim to substantiate in practice my goals of polymathy, in that I not only practice science and engineering, write, compose, play music and paint, but I devotedly and passionately engage all manners of physical activity, chiefly martial and practical in approach.

When one does not waste his time on stupid trivial pursuits, years to achieve set goals count fewer, as results manifest more abundantly and quickly, but if you are nonetheless interested in my age, it most likely means you are a woman, so I am determinately obliged to inform you that I am 12720.2 days old, experienced, motorically fully functional and abundantly technically versed, and I will refrain neither from jokes with sexual inuendo nor from personal metaphysical reflections on having quelled the vagina dentata.

Any further verbosity would no doubt produce invidious dismissal in even the most patient of readers, so I must converge on the final remarks. The kernel of the soul has already been alluded to; symbolic simulator revealed to be a simulacrum itself. Maladaptive engineering humor aside, if nothing else, the text ought to reveal a pursuit of truth, understanding and integrity, be it moral, personal, or scientific. And, of course, there are myriad others, equally deep and important, but they cannot and should not be revealed through a self-referential parody of an introduction, and those, dear respected reader, if you find a bit of yourself among these few paragraphs, I hope we can discuss in person.

Software development

I'm a former university professor with a Ph.D. in machine learning. Although the word career makes me bristle, I’m driven to keep pushing whatever environment I'm in toward something better. My primary interest is intelligence and, consequently AI and machine learning, largely because peering into neural networks illuminates how humans—and systems in general—learn. I’ve worked with a wide range of ML methods, and my dissertation explored this terrain in depth.

I've written extensively about intelligence in general and have made some strong and relatively unorthodox conclusions about the direction of the future development of software, architectures, AI, and intelligence in general. I released a detailed 100-page document (no AI-generated content) outlining some of these findings, which may serve as an additional source of credentialization and background to any interested party, aside from providing novel insights into intelligence and AI: Engineering Intelligence, Minds and Cognition. Of course, you are invited to read the rest of my writing, as well: Writing and Philosophy.

Despite appearances, I haven’t frozen into an “optimal configuration.” I’m always open to work that makes sense for both sides. At present I run an independent consulting practice, delivering AI expertise to multiple business clients, including as a Distinguished Technical Fellow at HTEC, where I previously served as Head of Artificial Intelligence. My role spans hands-on development and high-level AI strategy across multiple industries. Use the footer links if you’d like to talk.

C# and C++ suit me, but I’m fundamentally agnostic and pick tools to fit the problem. I’m comfortable with Python and JavaScript—though I dispute some of their design choices—and have shipped code in Java, C, and assembly. Learning new languages on demand is not an issue. Over 21 years I’ve built commercial products as both teammate and soloist, from intricate algorithms to old-fashioned puzzles. Below is a small, unordered sample of that work.

  • From-scratch AI architecture for a multi-billion-dollar retail giant including: AI agent coordination, automated customer support, security and policy, guidelines, internal source control, protocol implementation (including MCP and A2A), ethics guidelines and monitoring
  • General learning platform based on procedural interface generation and multi-agent lesson planning (GPT-4o, o4-mini, Claude 4 Sonnet, Gemini 2.5 Pro in the backend)
  • Conversational-AI agent cluster combining GPT-4 with cloud-hosted open-source LLMs—multiple coupled agents in a DERA-like arrangement to elicit emergent intelligence (some experiments nod to Jungian psychoanalysis)
  • Road-vehicle tracker and detector for surveillance (Python / TensorFlow, modified SSD network)
  • Transfer-learning model for aerial image classification (UCMerced + NWPU-RESISC45), published in IEEE GRSL with state-of-the-art results
  • Automatic polyp-detection algorithm for colonoscopy, published in Computers in Biology and Medicine
  • Procedural-music composition suite (C#): MIDI tooling, mini-DAW with VST host, audio renderer/converter, text-to-music description engine, style-compiler, and supporting upload utilities
  • Experimental compiler for a generalized procedural-generation language, specified with automata theory
  • Automatic method memoizer—runtime IL emitter that adds memoization to arbitrary C# methods
  • Full-stack school website (ASP.NET Web Forms, C#, MSSQL) with custom CMS, PDF generation, video transcoding, and image conversion—built before ASP.NET MVC matured
  • GPU-parallel many-body and particle simulations (OpenCL / CUDA; rendered via OpenGL or DirectX)
  • Custom CNN framework accelerated on GPU (OpenCL / CUDA back end, C# API)
  • Text-compression algorithm (C++) blending Huffman and LZW—outperforming RAR and ZIP for messages; shelved when priorities shifted
  • Virtualized automatic exam-scoring system (C# + assembly) with Windows–Linux interop, inspired by the IOI platform
  • Minimal operating system with its own shell and user-mode API
  • Content and video crawlers (C# server, JavaScript client) using regex pipelines to output XML—built to spare me from rewriting scrape scripts yet again
  • Dungeons & Dragons utility suite—UI/UX styled to match the tabletop aesthetic, complete with automated monster-image retrieval and proper artist attribution
  • VST equalizer with pitch-aware filters (C++, Steinberg SDK)
  • Music-training suite: audio engine, on-screen piano, tuner, and spectrometer (UWP, C#; FFT analysis and AMDF-based pitch detection)
  • Writer’s word-management tool—graph UI for definitions, synonyms, collocations, rhymes, portmanteaux (WPF)
  • Sierpinski-gasket enumeration on a supercomputer (C++, OpenMP)
  • Procedural Raven’s-matrices IQ-test generator with algorithmic puzzle logic and auto-generated UI visuals (UWP, C#)
  • Nutritionist suite (Delphi, 2008) that earned a silver innovation medal
  • Budget-optimized web-app implementing a spaced-repetition algorithm (Azure Storage Tables back end)
  • More than 100 apps across Windows and Android—everything from stopwatches to domain-specific platforms
  • Several Unity-based games (some VR-enabled on Oculus DK2, one featuring subcutaneous-tissue simulation)
  • North of 1000 projects overall—tools, research prototypes, full software suites, and the inevitable experiments
  • Occasional tongue-in-cheek “questionable” utilities—harmless pranks while actual malware was developed for learning

For the curious, I’ve spent time with MATLAB, Maple, Mathematica, Microsoft Office (plus custom add-ins), Adobe Creative Suite (ditto), and Cubase (VST development). Other technologies in shorter bursts include OpenMP, Hadoop, PHP / Yii, WordPress, PayPal API, F#, Haskell, Android, and MySQL—plus several I’ve happily forgotten (Spring among them).

I’ve set up and managed Azure web apps, VMs, and networks, and I live in Git. DevOps duties—CI/CD, deployment pipelines, automated tests—have been regular assignments.

Roles have ranged from systems architect and algorithm designer to full-stack developer and team lead (before that title fell out of fashion). I also interview and design assessments, a by-product of years teaching high-school physics, coaching programming Olympiads, and lecturing at university.

In industry I’ve spent 11 years building and refining ML systems with TensorFlow, PyTorch, Caffe, Keras, and fast.ai. My focus sits at the intersection of transfer learning and multi-task learning for classification, detection, and segmentation. I’ve designed bespoke architectures, tuned standard ones, and explored most auto-encoder variants; natural-language work is lighter but present.

These days, I advise organizations on AI, agentic development and agentic process automation. While serving as Head of Technology, Artificial Intelligence at HTEC I've built entire architectures end-to-end—from infrastructure and ethics to deployment and monitoring—for multi-billion-dollar enterprises. Past projects include ML-based super-resolution (that was likely running on your hardware at some point) and generative-AI-driven RPA.

Recent work ranges from agentic systems that exhibit emergent behavior to enterprise-scale RPA platforms lined with ML. I have, for example, architected a complete AI ecosystem for a major Middle Eastern retail conglomerate, delivering everything required for large-scale deployment.

Music, writing and art

Though informally trained, I’ve spent over 20 years studying composition and practicing several instruments. I gravitate toward pre-20th-century classical and Romantic music, which, to my ear, marks a high-water line of craft and expression. Ideological posturing in art leaves me cold, but I still explore plenty outside that era. I’ve written more than 60 pieces—mostly contemporary-classical and film-style scores—yet hesitate to publish until they meet the rather unreasonable standards set by my heroes.

Piano is my main vehicle for composing and improvising. I dabble in violin, flute, trombone, horn, guitar, and cello—mostly to hear how ideas live in real instruments—so proficiency varies. I’ve built an audio-synthesis engine and a handful of VSTs, but orchestral color remains my preferred palette, often blended with carefully chosen electronic timbres.

My cross-disciplinary itch led me to “meta-concepts” shared among fields. One product of that curiosity is an autonomous composition algorithm: a platform that turns abstract musical specifications into audio and MIDI, complete with a mini-DAW, VST host, custom compositional language, and even its own compiler.

Creating it forced me to digest mountains of music theory and translate the useful parts into mathematics. That process convinced me that standard theory is often descriptive rather than generative, and rarely aligned with what the ear craves. By melding software engineering, math, and music, I arrived at a more workable model—one that also sharpened my own musicianship.

I moonlight in photography, video, and digital painting, selling the occasional stock item. Large canvases demand time I rarely have, so many ideas remain sketches and studies—valuable mainly as personal experiments.

Ever the aspiring polymath, I keep philosophical journals, write short stories (allegorical and otherwise), and have several books languishing in pre-publication limbo in both English and Serbian.

Martial arts and physical arts

Socrates was right about nurturing body and mind alike, and I took that to heart long before I could articulate it. The stereotype of the frail academic irks me almost as much as its frequent accuracy, so I’ve chased martial arts, parkour, and solo sports with equal zeal.

A balanced body feels instinctively correct, despite modern culture’s attempts to argue otherwise. Athleticism shapes our sense of beauty and elegance, and martial arts, in particular, reveal what the human frame can do.

As an undergraduate I spent a few years in Krav Maga before realizing it had drifted toward monetizing insecurity. I migrated to Brazilian Jiu-Jitsu and now borrow liberally from BJJ, judo, boxing, and wrestling, with occasional inputs from kali, pankration, and the more traditional styles. The goal is a coherent MMA-flavored system that maximizes athletic potential—the “arts” in martial arts.

Traditional forms such as Wing Chun have dwindling practical utility, yet their drills still hide useful mechanics if you dig. The kinetic chain that powers a tai-chi strike, for instance, maps neatly onto the snap of a boxing punch. Discovering that the motion of drawing a violin bow mirrors the uppercut’s engine is a perfect example of a meta-concept—principles that cross domains and accelerate learning everywhere else.

I ground my practice in kinesiology and biomechanics—physics in motion. After 12 years training alongside athletes from many disciplines, I recently earned a Brazilian Jiu-Jitsu blue belt under Đorđe Vasić. The belt mainly clears the way to open my own academy; more on that at Gavran BJJ.