👋 Hi, I'm

MD Millat Hosen

AI Researcher · Educator · Tech Content Creator

I research LLM fine-tuning in resource-constrained settings and ship practical tools for developers — then teach both on MILLATechCode, a Bangla + English YouTube channel on Claude Code, AI agents, and the tools I actually use.

01. About

I'm a B.Tech CSE student at Sharda University (graduating August 2026), focused on AI research and developer education.

My work spans AI research, software development, and tech education. I specialize in fine-tuning large language models (Mistral-7B) using parameter-efficient techniques like LoRA and QLoRA, with a focus on deployment in resource-constrained settings. I've authored two arXiv preprints —arXiv:2504.15610 on parameter-efficient fine-tuning for educational guidance (with 5 citations and 1 Highly Influential Citation on Google Scholar) and arXiv:2606.21429 (co-authored with MN Sakib) on student migration patterns from Bangladesh to NCR private universities.

Beyond research, I run MILLATechCode, an independent YouTube channel where I test and review the AI tools I actually use — Claude Code, AI coding agents, and the latest open-weights releases — in Bangla with English auto-dub. Previously, I was a Digital Marketing & SEO Intern with Sharda's International Relations Division and a Web Developer at SIAC Abroad.

02. Research Interests

LLM Optimization

  • • Parameter-efficient fine-tuning (LoRA, QLoRA)
  • • Model compression for edge deployment
  • • Resource-constrained AI systems
  • • Efficient inference optimization

Autonomous Agents

  • • Multi-agent collaboration architectures
  • • Memory systems & context persistence
  • • Agent-based reasoning & planning
  • • Tool use & API integration

RAG & Retrieval

  • • Retrieval-augmented generation systems
  • • Vector databases & semantic search
  • • Contextual chunking strategies
  • • Hybrid retrieval methods

AI for Education

  • • Personalized learning systems
  • • Educational AI for developing regions
  • • Scalable EdTech platforms
  • • Adaptive assessment systems

03. Writing

04. Experience

Independent Tech Content Creator

MILLATechCode — Self-employed

Jun 2026 - Present · Remote

  • • Produce Bangla + English AI-coding tutorials and tool reviews on YouTube — Claude Code, AI agents, and open-weights releases
  • • Manage the full pipeline: research, scripting, recording, publishing, SEO, and audience growth
  • • Test sponsored tools with full editorial independence and disclose all partnerships transparently

Digital Marketing & SEO Intern

Sharda University — International Relations Division

May 2025 - May 2026 · Greater Noida, India

  • • Supported digital outreach to strengthen the university's global presence
  • • Managed the official Study at Sharda University YouTube channel for BD-facing admissions content
  • • Ran SEO and keyword research to improve international student recruitment campaigns
  • • Collaborated with internal teams on social-media planning and performance analysis

Web Developer

Study International Admission Care (SIAC)

Feb 2023 - Mar 2024 · Hybrid

  • • Built and launched the production website (siacabroad.com)
  • • Led marketing initiatives across Facebook, Google, and YouTube
  • • Maintained and optimized the web infrastructure

05. Technical Skills

AI / ML

PyTorchHugging FaceTransformersLoRAQLoRALangChainVector DBsMCP

Development

TypeScriptPythonReactNext.jsNode.jsTailwindSQLGit

Infrastructure

LinuxVercelDockerPostgreSQLMongoDBNginx

06. Featured Projects

Kitovo.app →

Free Receipt Generator Tool

Professional receipt generator tool with 47+ templates (rent, gas, taxi, donation receipts). 100% private (client-side processing), works offline, exports to PDF/PNG at 300 DPI. Built for small businesses, freelancers, and landlords. Features custom branding, automatic calculations, and draft saving. Achieved 90% keyword coverage and 439ms avg load time.

🚀 72 pages deployed⚡ 650KB bundle optimized🔒 100% client-side privacy
Next.jsReactTypeScriptSEOCloudflare Pages

A LoRA-Based Approach to Fine-Tuning LLMs →

Published Research • arXiv Preprint

arXiv:2504.15610 — Parameter-efficient fine-tuning of Mistral-7B-Instruct for academic advising in study-abroad contexts. Achieved 52.7% reduction in training loss, 92% accuracy in domain-specific recommendations, and 95% markdown formatting support. Published April 2025, indexed on NASA ADS, Semantic Scholar (5 citations, 1 Highly Influential), Google Scholar, ResearchGate, and University of Warwick. Preprint only — open to collaborations on a systems-flavored revision.

📄 18 pages, 6 figures📚 5 citations (1 Highly Influential)🎓 Indexed on 7+ platforms🔬 arXiv preprint
LoRAQLoRAMistral-7B4-bit QuantizationEducational AILow-resource ML

MILLATechCode →

YouTube Channel · Bangla + English · UCYnFScY04p-iXoJiKjhP2cg

Independent YouTube channel on practical AI coding: Claude Code, AI agents, free API access, and open-weights releases — tested hands-on before each review. Bangla as the recording language with English auto-dub for global reach.

📈 222 subscribers👁 24k+ views🎬 59 videos🤝 Sponsored reviews with full editorial independence
Content ProductionOBSCapCutAI ToolingSEOBanglaEnglish

Push-Pull Determinants in NCR Private Universities →

Co-authored Research · arXiv Preprint (with MN Sakib)

arXiv:2606.21429 — Single-destination exploratory study on what pulls Bangladeshi students to NCR private universities and what pushes them away. Co-authored with MN Sakib. Preprint only.

📄 arXiv preprint👥 Co-authored with MN Sakib🎓 Higher-education research
Higher EducationStudent MigrationBangladeshIndia

Message Thread Cleaner →

Browser Extension · Privacy-first chat management

Browser extension for cleaning and organizing message threads. Built with privacy as the primary constraint — all processing happens locally, no data leaves the device.

🔒 100% client-side🧩 Chrome Web Store📄 Dedicated privacy policy
Browser ExtensionTypeScriptPrivacy-firstLocal Processing

Open Source Portfolio →

29+ Public Repositories

A growing collection of public repositories — MCP servers, AI tooling, browser utilities, and experimental scripts. Range from working production code to learning projects.

📦 29+ repositories⭐ Active contributor🤝 Includes gsc-bing-mcp and BDLawCorpus
TypeScriptPythonReactAI/MLOpen Source

07. Sponsors & Partners

Brands and platforms that have supported MILLATechCode through paid collaborations, model access, or platform partnerships. Every sponsored review is published with full editorial independence and disclosed in the video and description.

Full sponsorship disclosure on every video. All partnerships are paid; no company reviews or approves content before publication.

08. Education

Bachelor of Technology (B.Tech.) in Computer Science & Engineering

Sharda University, Greater Noida, India

September 2022 - August 2026 (Expected)

  • • Focus: AI/ML, LLM fine-tuning, and software engineering
  • • Verified academic email at ug.sharda.ac.in (Google Scholar)
  • • Relevant coursework: Machine Learning, Deep Learning, Algorithms, Data Structures
  • • Two arXiv preprints: LoRA-based fine-tuning for educational guidance (5 citations, 1 Highly Influential) and a 2026 study on Bangladeshi students in NCR private universities

Higher Secondary Certificate (HSC) - Science

Aftabgonj Govt College

2019 - 2021

09. Get in Touch

Open to research collaborations, sponsored reviews, and partnerships that respect editorial independence. I respond fastest to email and sponsor inquiries through LinkedIn.