Behrang Mehrparvar

(AI Futurist * Cognitive Scientist * Innovation Leader * AI Researcher)

[LinkedIn] [Resume (academic - innovation)] [Publications] [Weblog] [Artworks] [Facebook] [Email]

Overview
Behrang Mehrparvar is an interdisciplinary researcher and engineer working at the nexus of generative AI, computational cognitive modelling, and the philosophy of mind. He combines deep technical expertise in neural networks, language models and applied machine learning with longer-term, theory-driven interests in representation, associative memory, Hebbian learning, and interbrain synchrony. Behrang holds a Ph.D. in Computer Science (University of Houston) and is completing an M.Sc. in Brain & Cognitive Sciences at the University of Amsterdam; he is also the founder of Synaptosearch, a lab/initiative focused on AI-driven approaches to representation and human-AI scaffolding.

Professional trajectory & formative projects

Behrang's interest in computation began in childhood and matured through progressively more ambitious projects: early algorithm and game development in QBasic and Assembly, state-machine hardware projects (a Morse transmitter) and AI-oriented software such as expert systems and minimax-based games. These formative projects gave him both low-level systems intuition and a habit of building complete, end-to-end systems—an engineer's sensibility he retains alongside his research ambitions.

During his Ph.D. (University of Houston) Behrang pursued ideas that explore why deep networks work—studies in community analysis of deep networks and conceptual domain adaptation—anticipating questions of representation and disentanglement that remain central to modern interpretability research. His early academic publications include a conceptual domain adaptation paper (arXiv). [link]

Industry research & product experience

After his doctorate Behrang accumulated six years of applied research and engineering experience in industry roles where he translated research ideas into production-oriented solutions. Notable responsibilities included designing transformer-based, generative solutions and reinforcement-learning components for search and query processing, building semantic-similarity graphs for query reduction/expansion, proposing a two-step fast spell-checking pipeline based on vector representations, and applying RL/GA/GAN techniques for web-application security assessment. These roles sharpened his ability to ship robust systems while preserving scientific rigour.

Recent academic work & collaborations

Since returning to academic research, Behrang's recent projects span multilingual ambiguity detection with LLMs, interbrain synchrony studies, and theoretical work on mental continuity and philosophy of mind.

• He co-authored “Detecting and Translating Language Ambiguity with Multilingual LLMs” (MRL 2024), a study showing how translation model representations can be used to detect ambiguity and how multilingual translation can be leveraged to preserve or reveal ambiguous meanings. [link]
• He is leading/participating in work on interbrain synchrony during collaborative decision-making and proposals for neurofeedback frameworks that scaffold synchrony in teams—projects that combine experimental cognitive neuroscience methods with computational modelling.
• He continues to write and publish reflections at the intersection of consciousness, synchronicity and AI. [link] [link]

Founding Synaptosearch & independent initiatives

Through Synaptosearch (an independent research platform he founded), Behrang pursues projects that sit between product R&D and exploratory science: proposals for an AI-optimized global language, associative-memory based matching engines, and generalized adaptive scaffolding AIs that support human learning and creativity. Synaptosearch is the vehicle he uses to iterate prototypes that embody his theoretical commitments (association, Hebbian mechanisms, representational clarity) and to demonstrate applied value.

Technical skills, methods & tooling

• Core ML / DL: Transformer architectures, autoregressive and encoder-decoder language models, CNNs, RNNs, GANs; generative modelling for structured content.
• ML engineering & libs: PyTorch, TensorFlow, Keras, scikit-learn; strong Python ecosystem skills (NumPy, pandas).
• Research methods: Representation analysis, domain adaptation, interpretability/analysis of hidden representations, experimental design for cognitive/neuroscience studies (interbrain synchrony).
• Systems & algorithms: Practical experience building search/vectorization systems, RL agents for security assessments, and novel spell-checking/vector update cycles.

Publications & selected outputs (highlights)

• Detecting and Translating Language Ambiguity with Multilingual LLMs, MRL/ACL Workshop (2024) — empirical and representational work on ambiguity via translation models. [link]
• Conceptual Domain Adaptation Using Deep Learning (arXiv, 2018) — work on domain adaptation framed around conceptual representations. [link]
• Consciousness: war, and peace! (PsyArXiv preprint, 2025) — interdisciplinary essay linking philosophy of mind and social issues. [link]
• Earlier proceedings on model-checking and formal methods (FCS 2011) and additional technical reports from industry projects are listed in his full CV.

Teaching, leadership & collaboration

Behrang has several years of formal teaching and TA experience (e.g., Advanced Machine Learning, Data Structures) and has supervised/mentored students in project-based settings. He is comfortable leading both small research teams and interdisciplinary collaborations that combine neuroscience, computation and engineering.

Intellectual stance & unique strengths

• Representation-first thinking: a persistent belief that representation (disentanglement, associative structures, memory primitives) is the central lever for progress in AI.
• Bridges research ↔ product: demonstrable experience turning theoretical ideas into applied systems (search, query processing, security tools).
• Interdisciplinary fluency: comfortable navigating philosophy of mind, cognitive neuroscience and machine learning—valuable for labs or companies building human-centric AI.
• Original thinking & foresight: a track record of pursuing non-fashionable but foundational directions (e.g., early advocacy for deep nets in his PhD, long-term focus on associative memory).

Career goals & how he adds value to teams

Behrang seeks roles that combine rigorous research with product impact: research engineer / applied research scientist roles in generative systems or human-AI interaction; interdisciplinary academic posts that bridge computation and cognition; or leadership roles in R&D groups exploring novel representation primitives and neuro-computational scaffolding. He contributes immediately through prototype development, experimental design, representational analysis, and by offering a strategic, long-horizon perspective on where AI should invest next.

[LinkedIn] [Resume] [Publications] [Weblog] [Artworks] [Facebook] [Email]

Last Update: 14 September 2025