Open to AI/ML Engineer & Research Engineer roles

Hi, I'm Meherab Hossain Talukder

AI / Machine Learning Engineer · Computer Science & Engineering Undergraduate

I turn research papers and messy data into AI systems that actually ship.

3× National Runner-Up11+ AI/ML Projects
View My WorkDownload Resume

About

Building AI that ships, not just notebooks

Computer Science & Engineering undergraduate (CGPA 3.52/4.00) building end-to-end AI systems across computer vision, NLP/LLMs, and time-series forecasting — First Runner-Up at three national AI competitions. Experienced across the full ML lifecycle, from large-scale feature engineering and gradient-boosted/deep models to LLM/RAG pipelines and FastAPI deployment. Targeting AI/ML Engineer, Data Scientist, and Research Engineer roles.

My work usually starts the same way: pick a problem that actually matters — churn, traffic safety, air quality, medical triage — and push a model past the notebook into something with an API, a UI, or a live inference pipeline behind it. I'm currently finishing my undergraduate thesis on PM2.5 forecasting for Dhaka, competing in national AI hackathons and datathons on the side, and looking for an AI/ML Engineer, Data Scientist, or Research Engineer role where I can keep doing both.

11+
Projects Shipped
3
National Runner-Up Wins
3.52/4.00
CGPA
CV · NLP · MLOps
Focus Areas

Projects

Selected Work

A mix of competition-winning models, research projects, and full-stack systems — spanning computer vision, NLP/LLMs, and time-series forecasting.

Customer Churn Prediction Using Survival Analysis

Code

bKash x NSUCEC Cybernauts 2026 Datathon Grand Finale — First Runner-Up

  • Reframed churn as a time-to-event problem and engineered 152 behavioral features from millions of transaction records
  • Calibrated a LightGBM + XGBoost + CatBoost ensemble with rigorous cross-validation to score churn risk and time-to-churn
PythonLightGBMXGBoostCatBoostScikit-learnPandas

RoadVision — AI-Powered Traffic Detection & Vehicle Classification

Code

DUET CSE Carnival 2026 AI Hackathon — First Runner-Up

  • Trained a YOLO11 (Ultralytics) detector for multi-class vehicle identification on Bangladesh highway surveillance footage
  • Built the augmentation and hyperparameter-tuning pipeline; delivered real-time inference evaluated with mAP@50/50–95
PythonPyTorchYOLO11UltralyticsOpenCV

Starship Safety — System Failure Prediction

Code

Data Hackathon, 5th National Data Science Summit (NDSS) 2026, Daffodil International University — First Runner-Up

  • Recovered hidden patterns from 15 obfuscated features through hex decoding, temperature parsing, and coordinate analysis
  • Achieved 94.5% accuracy / 0.944 F1 (cross-validated) with an interpretable logistic-regression + ColumnTransformer pipeline
PythonScikit-learnPandasNumPy

Multi-Agent Medical Assistant — RAG, Vision & Literature Search

Code

Personal Project

  • Orchestrated four specialized agents with LangChain — medical image analysis (brain tumor, chest X-ray, skin lesion), document RAG, PubMed/web search, and safety guardrails
  • Implemented semantic retrieval over medical documents with Qdrant; Dockerized the service with GitHub Actions CI
PythonLangChainLLMsRAGQdrantComputer VisionDockerGitHub Actions

Bengali LLM Hallucination Detection

Code

Kaggle Competition

  • Designed a macro-F1-optimized hybrid pipeline combining rule-based validation, Bengali Wikipedia retrieval, NLI scoring (mDeBERTa), and semantic similarity (LaBSE)
  • Integrated quantized LLM judges (Qwen2.5-7B 4-bit, Gemma via llama.cpp) for retrieval-augmented verification
PythonPyTorchHugging Face Transformersllama.cppRapidFuzz

PM2.5 Air Quality Forecasting for Dhaka

Code

Undergraduate Thesis (Ongoing)

  • Benchmarked LSTM, GRU, and ARIMA vs LightGBM, XGBoost, CatBoost for 24-hour-ahead forecasting on 2016–2022 data
  • Reached R² = 0.88 (LightGBM benchmark) using chronological splits, 24-hour-mean targets, and SHAP analysis
PythonTensorFlowLightGBMSHAPTime-Series Analysis

FinNova — Financial Transaction Risk Detection

Code

Personal Project

  • Detected high-risk transactions under severe class imbalance with a threshold-optimized weighted ensemble of LightGBM, XGBoost, CatBoost, logistic regression, and a neural network
  • Achieved ROC-AUC 0.979, AUC-PR 0.891, and F1 0.895 via systematic EDA, feature engineering, and threshold tuning
PythonScikit-learnLightGBMXGBoostCatBoost

Office Energy Monitor — IoT Energy Intelligence Platform

Code

Personal Project

  • Built a real-time energy monitoring platform: FastAPI + SQLAlchemy backend, React + Chart.js dashboard, and WebSocket streaming from Wokwi-simulated ESP32 sensors
  • Added room-wise analytics and a Discord alert bot for high-usage, after-hours, and long-running-device events
PythonFastAPIReactWebSocketsSQLAlchemyESP32IoT

SasthoSetu — Healthcare Platform Backend

Code

Personal Project

  • FastAPI + PostgreSQL healthcare backend with JWT/OAuth2 authentication, symptom checking, and appointment scheduling
PythonFastAPIPostgreSQLJWTOAuth2

Vantage Robotics Simulator

Code

Personal Project

  • Browser-based robot-arm control with forward/inverse kinematics, built with React, TypeScript, and Three.js
ReactTypeScriptThree.jsRobotics

Cirrhosis Outcome Prediction

Code

Machine Learning Project

  • Multi-class clinical outcome classification with EDA, feature engineering, and stratified cross-validation
PythonLightGBMPandasScikit-learn

Skills

Technical Toolkit

Programming

PythonC++CJavaJavaScriptTypeScriptSQL

Machine Learning

Scikit-learnLightGBMXGBoostCatBoostEnsemble LearningSurvival AnalysisFeature EngineeringCross-ValidationImbalanced DataSHAP

Deep Learning

PyTorchTensorFlowCNNLSTMGRUTransfer Learning

Computer Vision

YOLO11UltralyticsOpenCVObject DetectionImage Classification

NLP & LLMs

Hugging Face TransformersLangChainRAGmDeBERTaLaBSEQwen2.5Gemmallama.cppPrompt Engineering

Backend & Data

FastAPIREST APIsWebSocketsSQLAlchemyJWT/OAuth2Node.jsPandasNumPyEDA

Databases

PostgreSQLMySQLSQLiteOracleSQL ServerQdrant

Tools, DevOps & IoT

GitGitHubDockerGitHub ActionsJupyterKaggleESP32/ArduinoWokwi

Achievements

Competitions & Recognition

First Runner-Up — bKash x NSUCEC Cybernauts 2026 Datathon Grand Finale

National datathon

First Runner-Up — Data Hackathon, 5th National Data Science Summit (NDSS) 2026

Daffodil International University

First Runner-Up — DUET CSE Carnival 2026 AI Hackathon

  • 39th National Science Fair — 1st Place (Sub-district & District)National-level participant
  • 40th National Science Fair — 1st Place (Sub-district)
  • Creative Talent Hunt 2019 (Mathematics & Computer) — 1st Place (Sub-district)
  • 41st National Science Olympiad — 3rd Place (Sub-district)
  • BUBT AI Conquest 2025Competitor
  • The Infinity AI BuildFest 2026CloudCamp Bangladesh x BRAC University
  • IUBAT Technocrate V2 (Techathon)Competitor
  • IUBAT Innovation Competition 2024Line Follower Robot & Project Showcase

Experience

Work & Leadership

  1. Trainee IT Executive

    CODETREE, Uttara, Dhaka · 3 months

    • Provided IT support and hardware/software troubleshooting; assisted with system maintenance and basic networking
  2. Robotics Mentor

    IUBAT IT Society, Robotics Wing

    • Mentored juniors in robotics fundamentals; guided teams through competition preparation
  3. Academic Mentor, C Programming

    IUBAT IT Society, Academic Wing

    • Led peer debugging and problem-solving sessions

Education

Academic Background

B.Sc. in Computer Science & Engineering

International University of Business Agriculture and Technology (IUBAT), Dhaka

Expected 2026 · CGPA 3.52/4.00

HSC (Science)

Govt. Nalchity College

2020 · GPA 4.50/5.00

SSC (Science)

Nandikathi Adarsha Secondary School

2018 · GPA 4.28/5.00

Certifications

  • Database Systems (MySQL, Oracle, SQL Server)

    EDGE Project, Bangladesh Computer Council, ICT Division · 6 months

  • The Ultimate Job Ready Data Science Course

    CodeWithHarry · 2025

  • Complete 2025 Python Bootcamp: Learn Python from Scratch

    CodeWithHarry · 2025

View all certificates →

Contact

Let's build something together

Have an internship, research collaboration, or role in mind? Send a message and I'll get back to you.