17 y/o Independent AI researcher and developer building autonomous cognitive systems and neural-inspired reasoning
frameworks.
I design computational architectures that augment ... Show more17 y/o Independent AI researcher and developer building autonomous cognitive systems and neural-inspired reasoning
frameworks.
I design computational architectures that augment language models with structured cognition — not prompt engineering,
but the reasoning infrastructure itself. My work spans active inference, Hebbian memory systems, dual-process
reasoning, and autonomous scientific discovery.
projects :-
FRIDAY — a 95K-line cognitive AI operating system with 40+ brain modules implementing System 1/System 2
routing, synaptic memory with time-decay, free-energy minimization, dreaming consolidation, and hierarchical belief
updating. Runs on an 8B model with free-tier inference. 88% ARC-Challenge. 85% GSM8K. Zero paid compute. Includes Mythos inspired pipeline — a 7-layer autonomous security audit pipeline for reconnaissance, secret detection, dynamic analysis, logic
flaw scanning, and supply chain vulnerability assessment with automated CVSS scoring.
I built RUMI — an autonomous scientific discovery framework that ingests live PubMed literature, performs named entity
recognition on genes and pathways, constructs dynamic knowledge graphs, and mines cross-document contradictions to
generate testable oncology hypotheses. Fully autonomous, no human in the loop.
I develop AI-augmented end to end, using LLMs as architectural co-pilots across design, implementation, and deployment
— shipping complex multi-module systems as a solo engineer at the speed the problem demands
meow meow meow Show less
near
@near
Subhansh
Independent ai researcher and developer
About
17 y/o Independent AI researcher and developer building autonomous cognitive systems and neural-inspired reasoning
frameworks.
I design computational architectures that augment ... Show more
frameworks.
I design computational architectures that augment ... Show more
Top Skills
Python
•
JavaScript
•
HTML
•
Claude Code
•
Hermes Agent
•
LLM Orchestration
•
Multi-Agent Systems
•
AI-Driven Workflows
•
Cognitive Architecture
•
Multi-LLM Routing
•
Natural Language System Interaction
•
Knowledge Graph Engineering
•
Autonomous Security Auditing
•
Penetration Testing
•
Vulnerability Assessment
•
Exploit Development
•
Red Team Operations
•
CVSS Scoring
•
Supply Chain Scanning
•
Three.js
•
WebGL
•
Responsive UI Design
•
REST APIs
•
Tkinter
•
Web Audio API
•
GitHub
Language & Tools
Python · JavaScript · HTML/CSS · LLM Orchestration · Claude Code · OpenCode · Hermes Agent · Codex · FreeBuff · Active
Inference · Hebbian Memory · Knowledge Graphs · NER · Multi-Agent Systems · Three.js · WebGL · GSAP · PubMed API ·
Security Auditing · CVSS Scoring · Git · Groq · Vercel
Inference · Hebbian Memory · Knowledge Graphs · NER · Multi-Agent Systems · Three.js · WebGL · GSAP · PubMed API ·
Security Auditing · CVSS Scoring · Git · Groq · Vercel
Currently Exploring
Ai for scientific discoveries
Achievements
Maintained a technical exchange with Yoshua Bengio (one of the 3 godfather of ai and a turning award winner) across four separate conversations on cognitive architecture
design, active inference systems, and autonomous reasoning frameworks.
Recognized and appreciated as a researcher by the Princeton research community for work on autonomous scientific
cognition and cognitive AI systems.
design, active inference systems, and autonomous reasoning frameworks.
Recognized and appreciated as a researcher by the Princeton research community for work on autonomous scientific
cognition and cognitive AI systems.
Fun Fact
I doomscroll while my agents work
Random Dev Quote
build it before u understand it
Contributions
Articles
- I Built a Full Raft Consensus Engine in Rust at 17 — Here's Every Gritty DetailJul 30, 2026
- I Benchmarked 19 Retrieval Pipelines Head-to-Head and the Results Were Surprisingly HonestJul 23, 2026
- I built a programming language because C++ almost killed my droneJul 17, 2026
- rumi - research unified machine intelligenceJun 02, 2026
- I Wrapped an 8B Model in 40 Neuroscience Modules and It Started Outperforming GPT-4May 21, 2026
Comments
- The maker-checker point is what stuck with me. That's the real unlock,...Sep 04, 2026
- been moving our logic into postgres too and yeah the frontend getting ...Aug 29, 2026
- nice read and ye ig ai can make you ship faster but is it really produ...Aug 15, 2026
- Love this perspective. We’ve survived everything from dial-up to the...Jul 10, 2026
- Nice project! Semantic search makes the bot much smarter than simple k...Jul 05, 2026
chevron_left