// about

Hi, I'm Sohag.

I build the systems that run in the background so people don't have to.

SYSTEM MAP01 — 04
INPUT······························OUTCOME
Sohag Gain
// quick facts
Dhaka, Bangladesh · Remote (GMT+6)
BSc CSE · Jamalpur Science & Technology University · CGPA 3.55/4.00
English, Bengali, Hindi
Usually replies within 24h
// background

Sohag Gain is a Dhaka-based AI Automation Engineer and AI Agent Developer. As founder of AI Smart Galaxy, he has delivered 100+ projects for 25+ clients across CRM, AI agents, AI-enabled applications, and workflow automation. His work also spans AI Engineering and AI Solutions Architecture.

BSc in Computer Science & Engineering from Jamalpur Science & Technology University (CGPA 3.55/4.00). His thesis benchmarks machine learning and transformer architectures for Bangla–English news classification.

Read more about my background

That grounding is why my automation work isn't just drag-and-drop — I understand what's happening under the workflow builder.

// research foundation

Applied AI research, with the paper behind it

My BSc thesis evaluates classical machine learning, BiLSTM, BERT, and a hybrid LSTM+BERT approach across 562,163 Bangla and English news articles. Under the same evaluation pipeline, BERT reached 92.95% accuracy for Bangla and 94.50% for English classification.

See thesis scope and results

Thesis: Benchmarking Classical Machine Learning and Transformer Architectures for Bangla–English News Classification: A Large-Scale Unified Evaluation with Morphological Complexity Analysis. The work compares Logistic Regression, Naive Bayes, SVM, Random Forest, XGBoost, BiLSTM, BERT, and LSTM+BERT with an 80/10/10 stratified split. It also shows when a TF-IDF + SVM baseline can be a practical choice for lower-resource deployments.

Read the full thesis (PDF) → · View academic profile →

// what i do

Founder & CEO of AI Smart Galaxy — automation systems and AI agents for clients across the US, UK, UAE, and Bangladesh.

See how I work with clients

Public GitHub portfolio of working builds — I show a system running, not just describe it.

// how i think

I don't automate tasks. I redesign systems.

→Start with the bottleneck, not the tool.
→Map the process before building.
→Design for failure, not just the happy path.
→Use AI where it creates leverage.
→Keep systems maintainable.
→Build around the outcome, not the stack.
// how i work

Trigger → decision → action → human-readable log.

See the full workflow

Capture the trigger → let an AI agent or workflow decide → take the action → log it where a human can check. Simple, auditable, and built to run without me.

// currently learning

Building deeper skills in RAG, multi-agent systems, cloud deployment, and voice AI.

See what I'm learning now

Vector databases and RAG architecture, multi-agent frameworks like LangGraph and CrewAI, cloud deployment (Docker, Linux, VPS), and AI voice/video tooling. Not claiming mastery of all of it yet — just showing the direction I'm building toward.

// values
Impact
Growth
Leadership
Innovation
Continuous Learning
Integrity
Long-Term Thinking

Want the technical details?

See what I've actually built, or grab a copy of my resume.

View Portfolio See Resume Academic Profile