Available · full-time

Md Karimul Islam

Software Engineer

AI / ML Engineer · Computer Vision · Explainable AI

CSE graduate (CGPA 3.90/4.00) with hands-on experience building backend services, APIs, and database-backed data pipelines — shipping five end-to-end systems with clean, maintainable code. Comfortable across the stack, from data engineering to production deployment, backed by three first-author AI research publications and one co-authored.

Md Karimul Islam
5
Live AI projects
4
Publications
3.90/4.00
CGPA · B.Sc. CSE, Varendra University

What I've shipped

Live Projects

Five deployed apps spanning research, text, vision, audio, and production ML — each with a live demo and open-source code.

LitePhospho — Explainable Phosphosite Predictor

The live demo of my first-author thesis model: paste a protein sequence to predict phosphorylation sites (S/T/Y), each explained by the residues that drove it. My trained model, served via ONNX.

BioinformaticsExplainable AIONNXCNNStreamlit

Agentic Research Assistant

An LLM agent that answers questions across your research papers with page-level citations — then scores its own answers for faithfulness and relevance.

RAGAI AgentsLLM EvalFAISSONNXGroq

Multimodal Visual Analyst

Upload any image to ask questions about it or get a one-click structured JSON analysis — powered by a vision-language model.

MultimodalVLMLlama 4 ScoutStructured Extraction

Audio Intelligence

Turn a meeting or lecture recording into a transcript, a structured report (summary, action items), and a Q&A chat — local Whisper + LLM.

Speech-to-TextWhisperLLMStreamlit

Churn Prediction — MLOps

A model I train (scikit-learn) served in a production way: FastAPI + validation, a customer form, Prometheus metrics, a live dashboard, Docker and CI.

MLOpsFastAPIscikit-learnDockerCI/CDMonitoring

Research

Publications

First author on multiple peer-reviewed AI works in computer vision and explainable AI.

MEFNet — Hybrid CNN for Waste Classification + XAI

2026

First Author · IEEE PECCII 2026

MobileNetV3 + EfficientNetB0 fusion; 97.71% accuracy across 12 categories with Grad-CAM interpretability — outperformed ResNet50V2, EfficientNetV2, ConvNeXt-Tiny.

DOI: 10.1109/PECCII70991.2026.11662088

FusionNet — Deepfake Detection Framework

2025

First Author · VIJIR Vol.1(2) 2025 · UCICS

Hybrid DenseNet121/169; 91.22% accuracy, AUC 0.98, F1 92.56%; adversarial robustness tested against FGSM attacks.

DOI: 10.64296/vijir.v1i2.01

LitePhospho — XAI CNN for Phosphorylation Prediction

2025–Present

First Author · Undergraduate Thesis · Ongoing

Lightweight 1D-CNN with Integrated Gradients explainability; ONNX export for low-resource deployment; homology-aware benchmarking.

DOI: 10.5281/zenodo.21290440

Enhanced MobileNetV1 for Autism Detection

2025

Co-Author · IEEE QPAIN 2025

Co-authored an early autism-detection model using an enhanced MobileNetV1 architecture.

DOI: 10.1109/QPAIN66474.2025.11172046

Toolkit

Skills

Programming Languages

PythonC++CSQL

Databases

MySQLPostgreSQLSQLite

API Integration

REST API DesignOpenRouterOpenAIFastAPI Services

Data Structures & Algorithms

TreesGraphsPriority QueuesDynamic ProgrammingBFS/DFS

Data Preprocessing

Homology-aware Clustering (MMseqs2)Sequence-window ModellingData AugmentationNormalization

ML / DL

PyTorchTensorFlowKerasScikit-Learn

Architectures

CNNsDenseNetResNetMobileNetEfficientNet1D-CNN

CV / XAI

OpenCVMediaPipeGrad-CAMIntegrated GradientsLIMEONNX

LLM / GenAI

Prompt DesignChain-of-ThoughtRAGAI AgentsLLM Evaluation

Tools / Serving

VS CodeGitFastAPIDockerLinuxStreamlitTesseract OCR

Journey

Experience & Education

Prompt Engineering Intern

Jan – Apr 2026

Oleyn PTE Ltd. (Oleyn.ai) — Remote

  • Engineered structured prompts for production LLM pipelines across 3+ task categories using chain-of-thought and role-based strategies.
  • Benchmarked model outputs on reliability and quality metrics; milestones certified by the Co-Founder & CEO (Future Leaders Program, Batch 1).

Project Lead

Apr – Sep 2025

GeoChain — Rajshahi

  • Built a blockchain-ready property marketplace with identity verification workflows, securing funding and completing structured entrepreneurship training through the ICT Division & World Bank-backed UIHP Innovation Cohort.

Project Lead

Oct 2024 – Mar 2025

VetConnect — Rajshahi

  • Launched a digital veterinary platform prototype with GPS-based discovery, telemedicine, and e-commerce in 6 months, successfully securing a UIHP grant and completing structured innovation training.

B.Sc. in Computer Science & Engineering

Sep 2022 – Aug 2026

Varendra University, Rajshahi

CGPA 3.90 / 4.00 · Merit Scholarship (2022–2026) · Hult Prize Campus Team Leader (2026)

Certifications

Google AI Agents Intensive (5-Day)Machine Learning with PythonData Visualization with PythonGoogle Analytics

Let's talk

Get in touch

Open to AI/ML engineering roles — available full-time. The fastest way to reach me is WhatsApp, or ask my AI assistant anything using the chat bubble.