Applied ML Specialist · Vector Institute · Toronto

Building AI agents and efficient multimodal systems.

I build evaluation pipelines, multimodal data systems, and optimized GPU inference for foundation models.

View résumé Email me
GitHub LinkedIn Scholar
Portrait of Mohammed Saidul Islam

Latest NeurIPS 2026 E&D Track

FLAME

Fine-Grained Benchmark Generation for Comprehensive Evaluation of Foundation Models.

Paper

Accepted at NeurIPS 2026 Evaluations & Datasets Track

Now

Recent milestones

Recent publication and career updates.

Sep 2026 FLAME was accepted at NeurIPS 2026 E&D Track.
Aug 2026 OpenPMC-18M was accepted at MICCAI 2026.
Aug 2026 DSAgentBench was accepted at EMNLP 2026 Main.
View all news

Selected work

What I built, how it works, and the results.

MICCAI 2026 · Multimodal ML

Open-PMC-18M

Biomedical figure extraction and summarization for medical vision-language learning.

My role: Owned preprocessing and optimized batched vLLM inference for extraction and summaries.

Result: Supported OpenCLIP training with the resulting image-text pairs.

  • VLMs
  • vLLM
  • Data Pipelines
Paper

NeurIPS 2026 E&D Track

FLAME

Reference-grounded evaluation tasks for foundation models.

My role: Built OCR ingestion, designer/verifier agents, iterative repair, and semantic deduplication.

Result: Generated benchmarks for machine learning and finance.

  • OCR
  • Multi-agent Systems
  • Benchmark Generation
Paper

EACL 2026 Main · Post-training

RL-Text2Vis

Post-training for executable, reliable chart generation.

My role: Co-developed GRPO feedback from generated text, code execution, and rendered charts.

Result: Improved executability and chart quality over prompting and supervised baselines.

  • GRPO
  • Post-training
  • Code Generation
Paper Code
Explore all technical work

Career

Experience and education

Roles in applied ML research, engineering, and teaching.

2025-Present

Vector Institute

Associate Applied Machine Learning Specialist · Toronto

Foundation-model evaluation, multimodal data pipelines, and GPU inference.

2023-2025

Intelligent Visualization Lab, York University

Graduate Research Assistant · Toronto

Agent workflows, chart reasoning, and interactive dashboard evaluation.

2021-2023

Islamic University of Technology

Lecturer · Bangladesh

MSc Computer Science - York UniversityBSc Computer Science and Engineering - Islamic University of Technology
View full experience

Research

Selected publications

Recent work on AI agents, multimodal evaluation, data storytelling, and reliable generation.

  1. NeurIPS Accepted

    Fine-Grained Benchmark Generation for Comprehensive Evaluation of Foundation Models

    Mohammed Saidul Islam , Negin Baghbanzadeh , Farnaz Kohankhaki , Afshin Cheraghi , Ali Kore , Shayaan Mehdi , Elham Dolatabadi and Arash Afkanpour
    arXiv preprint arXiv:2605.188242026

    NeurIPS 2026 Evaluations and Datasets Track

    Read paper
  1. EMNLP Accepted

    DSAgentBench: Can Agents Automate End-to-End Data-Science Workflows in Real Computer Environments?

    Mizanur Rahman , Mohammed Saidul Islam , Ridwan Mahbub , Md Tahmid Rahman Laskar , Shafiq Joty and Enamul Hoque
    arXiv preprint arXiv:2608.103662026

    EMNLP 2026 Main

    Read paper Code
  1. EACL Conference

    Aligning Text, Code, and Vision: A Multi-Objective Reinforcement Learning Framework for Text-to-Visualization

    Mizanur Rahman , Mohammed Saidul Islam , Md Tahmid Rahman Laskar , Shafiq Joty and Enamul Hoque
    Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics2026
    Read paper PDF Code
All publications

Contact

Let’s talk about ML engineering.

Email me LinkedIn GitHub Scholar