Dhruv Dubey
AI Researcher · LLM Memory Architectures
Designing token-efficient conversational history management systems and robust NLP pipelines.
Current Focus
Actively job hunting for Systems Engineering & AI Research roles
Seeking research collaborations in LLM Memory & OS Architectures
Preparing applications for Masters in Computer Science (MS CS)
DSPM: Dynamic Semantic Patch Memory
Token-Efficient Long-Dialogue LLM Context Management via a 7-technique pipeline.
Mean CRR: 94.2%
Wilcoxon-validated (p=0.0431)
"No lab. No GPU. Just a problem worth solving."
Open Source
Technical Arsenal
AI & ML
Systems & Tools
Experience
HCDS Technologies
Project Trainee · Nov–Dec 2024
Awarded Trainee of the Month.
Chegg India
Technical Mentor · Jan–Apr 2023
200+ problems solved (DSA, DB, CN).
Quick Heal Foundation
Cybersecurity outreach, 10+ workshops, 1,500+ students.
Education
B.Tech IT
BIET, AKTU, Lucknow
Jul 2022 – Jun 2026
(First Division)
SSPM: Sparse Semantic Patch Memory
Token-Efficient Conversational History Management. 166 views, 106 downloads.
100% constraint preservation
LoCoMo benchmark validated
Semantic Patch Triage RL Env
Meta PyTorch Hackathon × Scaler (Apr 2026). Solo, 48hrs, HuggingFace Space live.
Score ~0.79
Vyom OS
AI-Integrated Operating System Architecture built from scratch.
Pie
Open-source contributor — LLM inference engine, Prof. Lin Zhong's Efficient Computing Lab, Yale
VibeServe
Open-source contributor — GPU-efficient LLM serving system, Prof. Baris Kasikci's lab, UW
Let's build something.
+91 7652056545