AI Engineer & ML Practitioner — Dhaka, BD

Building
Intelligent Systems.

I engineer production RAG pipelines, LLM-powered copilots, and computer vision systems. BSc CS, BRAC University — currently building AI Solutions at Softvence.

View Projects Resume

GitHub LinkedIn Website

AI Engineer

Shaanzee.

00 — About

Who I Am.

"I don't just wrap APIs around models. I build systems — retrieval pipelines that actually retrieve, chatbots that actually understand, and vision models that actually generalise."

— Mahinuzzaman Shaan

LocationDhaka, Bangladesh

Current RoleAI Engineer — Softvence

EducationBSc CSE — BRAC University, 2025 · 3.56/4.00

SpecialisationGenerative AI, RAG, LLMs, Computer Vision

Emailmahinuzzaman.shaan@gmail.com

Status✓ Open to Opportunities

01 — Experience

Where I've
Worked.

Mar 2026 — Present

Softvence

Dhaka, BD · Full-time

AI Engineer

Architecting production-grade Generative AI systems — from RAG pipelines and LLM-powered copilots to intelligent voice interfaces — for enterprise clients across multiple verticals.

Architected production RAG pipelines with hybrid vector search in Pinecone, improving retrieval accuracy by 25%.
Developed GenAI copilots and automation tools via LLM API integrations, cutting manual enterprise overhead by 40%.
Engineered AI Voice Chatbots using Whisper STT and ElevenLabs TTS, achieving sub-second response latency.
Designed multi-turn chatbots with advanced prompt engineering and context injection, boosting intent recognition by 20%.

PythonLangChainRAG PineconeWhisperFastAPIAWS

Nov 2025 — Mar 2026

Mastercourse Bangladesh

Dhaka, BD · Trainee

Data Scientist Trainee

Built end-to-end machine learning pipelines for computer vision and tabular data tasks, from data preprocessing through model deployment and experiment tracking.

Built a Computer Vision system with PyTorch achieving 94% accuracy via transfer learning and data augmentation.
Reduced ML iteration cycles by 30% through systematic hyperparameter tuning and experiment tracking.
Optimised collaborative filtering engines, improving tabular prediction baselines by 15% using DL ensembles.

PyTorchTensorFlowScikit-learn KerasPandasNumPy

02 — Work