01
Three-time founder
Three companies built since the age of seventeen, spanning education, manufacturing, applied R&D, and AI infrastructure. Together they served more than 500 customers and generated over $500,000 in revenue.
03Milestones & recognition
Achievements
Evidence over adjectives. The ventures, research, and results below are documented and, wherever possible, linked to their source.
3x
Companies founded since age 17
$500K+
Revenue generated across ventures
485
Citations across 5 publications
82.44%
Avg Recall@5 on MSR-VTT
17
Stanford AI/CS courses (audited)
3.89
M.S. GPA, Applied Data Science
- 02
First exit at 21
Founded an education-services company at seventeen and led it to a successful exit in 2016, after serving more than 500 customers.
- 03
Built and scaled a manufacturing company
Launched a textile manufacturer in Bangladesh in 2016 and scaled it from a small setup to a $100K+ business within five years, exporting to five-plus countries while leading production, operations, supply chain, and client delivery.
Manufacturing film - 04
Company-backed R&D lab
Built and led an applied research lab of seven inside the company, including two professors and graduate researchers, which produced five peer-reviewed publications, 485+ citations, and a pending machine-learning patent for industrial prediction.
- 05
Cross-domain research track record
Published in materials science and energy storage while developing patent work in machine learning and data science, consistently moving across hard domains and turning research into usable products.
- 06
Strong academic foundation
Studied Applied Chemistry and Chemical Engineering at one of Bangladesh's top universities (under 2% acceptance), then earned an M.S. in Applied Data Science at the University of San Diego with a 3.89 GPA.
- 07
Archived master's capstone
Selected among the strongest projects in the cohort and archived in the university's digital library.
View capstone - 08
Stanford AI & systems training
More than three years in Stanford's AI and computer-science ecosystem through audited and independent study, completing 17 courses across deep learning, NLP, reinforcement learning, transformers, ML systems, compound AI, and self-improving agents.
- 09
Stanford research benchmark
As part of a Stanford research project, built a benchmark to evaluate model and system behavior under controlled experimental conditions.
View benchmark - 10
Multimodal memory for AI agents
Built a multimodal memory system for AI agents in 2025, reaching 82.44% average Recall@5 on MSR-VTT in our testing, with 97.08% text/audio, 77.86% image, and 71.75% video retrieval.
avoslab.com - 11
AI-agent developer infrastructure
Created avos-dev-cli, an open-source, Git-aware memory layer for AI coding agents such as Claude Code, Cursor, and Codex, giving them repository and session memory to reason over codebase history, decisions, and in-flight work.
GitHub - 12
Current mission
At Avos AI, building the context, memory, encryption, and control plane for a world where billions, and eventually trillions, of AI agents operate across software, business, and human systems.