Team
A team at the intersection of neuroscience and AI. We spent 10+ years researching how the human brain encodes, stores, and retrieves memory — then engineered the same principles into a system that works.
Andrew Kislov
PhD in neuroscience candidate. 10 years studying the brain through fMRI and EEG. His research focuses on the neurobiology of decision making, using fMRI and EEG to predict behavior at the population level from neural signals. His work has been published in Q1 journals.
Andrew founded and led Emonomy, a US-based MarTech startup that connects media spend to brand outcomes before budgets are committed. Under his leadership, Emonomy:
- Generated revenue across 20 countries, including the USA, Mexico, Singapore and the Netherlands
- Worked with 40+ global brands, including Unilever, Toyota, P&G, Nestlé, Honda, Citi, Subaru, and Pandora
- Secured investment from 500 Global, a top-2 early-stage VC worldwide
Founded a non-profit philosophical club with chapters in Yerevan, Berlin, Barcelona, and Belgrade.
Artem Gadzhiev
Global Master of Public Health (longevity & personalized health) from Imperial College London.
Former Director of Product & Technology at one of Europe's largest banks, where he ran the back office of Global Markets — AI-driven trading, compliance automation, and a company-wide AI enablement program. Shipped 15+ AI, fintech, and Web3 products end-to-end:
- Led cross-functional teams of up to 300
- Drove 2,000% YoY platform adoption on a unified AI roadmap
- Cut manual compliance oversight by 60% with LLM workflow automation
MSc CS (Machine Learning) and BSc Econ & CS from MSU. 20+ years of independent research in neuroscience and human biology.
Founded 3 startups in AI + Web3 space in the past 3 years.
Vision
Today your context is scattered across platforms that own it. Your music preferences live in Spotify. Your relationships live in WhatsApp. Your work history lives in LinkedIn. None of it talks to each other, and none of it belongs to you.
We believe the future looks different. Every person will have a structured, portable representation of themselves — their own context layer. A rich, evolving map of who they are: what they know, what they care about, how they think, who matters to them.
This is the foundation for a new kind of human–AI interface. A living layer that knows you and acts accordingly. Less input, more understanding. And it is the foundation for a new kind of connector — between a person's data and agentic functions: agents read your context with your permission and rely on it when taking action in the real world.
And you control it. Choose which AI agents get access. Decide what a healthcare assistant knows versus a work copilot. Grant and revoke context the way you grant permissions — deliberately, with full ownership.
This changes the relationship between people and AI. Instead of starting from zero every time, or trusting a platform to remember you on your behalf, you bring your own context.
Synthius is the infrastructure for that future. We start with the hardest part — structured memory extraction that actually works — and build toward a world where your context is an asset you own, not a byproduct someone else monetizes.
Without false modesty —
this is the best personal memory system in the world.
A memory and personality layer for AI.
Synthius pulls your scattered context — chats, docs, voice, calendar, anything — and organizes it into six structured domains of a human persona. Any AI agent you grant access to reads that memory with your permission and acts as if it's always known you.
#1 on LoCoMo. Near-perfect recall at a fraction of the token cost.
On the standard long-conversation memory benchmark, Synthius beats every other memory system — and full-context baselines running the strongest frontier models. High accuracy, no fabrication, real-time retrieval.
Full-context replays the whole history every message. Synthius extracts once and retrieves cheap structured facts forever — ~5× fewer tokens at N=500, without giving up accuracy.
Brain-inspired architecture — six domains, two phases.
The brain doesn't use one memory system — it uses six. Synthius mirrors that: six parallel extractors write to six domain stores during ingestion. At chat time, a planner routes each query to just the relevant stores — fast, grounded, auditable.
Every source runs through the same pipeline. 17 formats parsed, chunked, and split across six parallel extractors. Every update is a reversible diff, so nothing is ever clobbered.
A planner picks which domains matter for the query. Only those stores are read. The answer comes back grounded, cited, and honest when evidence is missing.
Most memory systems are a single vector blob hoping similarity search lands on the right fact. That's why they hallucinate and break at scale. Here's what we do differently.
| Usual AI memory | Synthius memory |
|---|---|
| One flat vector blob | Six brain-inspired domains |
| Similarity search across everything | Category-aware retrieval by domain |
| Opaque embeddings, no confidence | Structured facts with confidence & source |
| Append-only, can't undo | Reversible diffs, fully auditable |
| All-or-nothing context share | Granular permission per domain |
| Hallucinates personal facts under pressure | 99.55% hallucinations resistance · zero fabrication |
| ~67% accuracy (Mem0, LoCoMo) | 94.4% accuracy (LoCoMo) |
You can read about our memory in detail in the paper.
Read the paper →Or explore our memory on an example — Dostoevsky's persona.
See the example →Our Services
Memory for your project
Give your AI product persistent, structured memory out of the box. Users get recognized across sessions — their preferences, history, and context travel with them, while you save 80%+ of tokens. No custom infra needed. Plug in Synthius-Mem and your product remembers.
Personal memory MCP
Your own memory layer that works across every AI tool you use — Claude, ChatGPT, Cursor, and anything else that supports MCP. One structured context, portable everywhere. The AI already knows you, regardless of which tool you open.
Synthius platform
Use your memory directly inside Synthius — a home for agents that already know you. A coach and therapist that understand your life, a news feed filtered by your profile and interests, a personal assistant that picks up mid-thought, plus custom agents you assemble for your own use-cases. One memory, many lives, all yours.
Example: Explore Dostoevsky
We fed Synthius the complete biographical record of Fyodor Dostoevsky — his letters, memoirs, novels, and biographies. The system extracted a fully structured persona across six cognitive domains plus a chronology of life events.
Raw source material
Six cognitive domains
Below is what came out. The same depth is available to any Synthius user.
Social circle
Turgenev, Belinsky, Nekrasov, his wives, his brother Mikhail — mapped relationships, conflicts, debts, literary rivalries.
Psychometrics
Epilepsy and its influence on his worldview, gambling compulsion, oscillation between faith and doubt, emotional intensity.
Preferences
Tea over coffee, St. Petersburg over Moscow, night writing sessions, specific newspapers, literary tastes and dislikes.
Work
Editorial work on Vremya, Epoch, and Diary of a Writer; his publishing deals, debts, contracts, and the arc of his literary career.
Experiences
Siberian exile, mock execution, St. Petersburg flood, European travels, gambling binges, epileptic seizures — life events with emotional valence.
Biography
Family tree, birth and death dates, residences, education, languages, religion, nationality — 774 atomic facts across 12 categories.