Kazi Rafat Haa Meem

I'm an AI researcher working at the intersection of speech and large language models. I build speech and audio systems (recognition, synthesis, and signal processing for real-world audio), agentic LLMs, and efficient deep learning through knowledge distillation and model compression, with a strong focus on deploying them in health-critical and multilingual settings. I go deep on architecture, training, and evaluation, and I care about models that are both state-of-the-art and genuinely deployable. First author at Interspeech 2026 and IEEE JBHI, with work in PLOS One and Pattern Recognition Letters, and three years shipping production ML at Decodis. Currently searching for the right PhD.

Now [Currently working on Audio Language Models and Speech to Speech Systems. Applying.]

📰 News

Apr 2026Selected for the Adaption Labs inaugural grant cohort (AdaptBench, LLM evaluation) — the new adaptive-AI lab founded by Sara Hooker.
decodis
Jan 2026Concluded three years at Decodis as a Machine Learning Engineer.
2026First-author streaming ASR paper accepted at Interspeech 2026.
2026Point Cloud Transformer for physical rehabilitation at IEEE JBHI.
PRL
2024DeepMarkerNet published in Pattern Recognition Letters (Elsevier).
2024Low-resource Bangla TTS presented at the ICLR Workshop on Practical ML.
2023Knowledge-distillation paper published in PLOS One; later featured at Cohere Labs.
2023MenoChat won Best Idea + Best Poster and a $30K grant at NCSRHR.
decodis
2023Joined Decodis (Boston, remote) as a Machine Learning Engineer, building multilingual speech and language systems.

📄 Publications

Dynamic Block-Online Streaming ASR for Low-Resource Agglutinative Code-Switching Speech with Morphology-Aware Evaluation

Interspeech 2026First author

Kazi Rafat Haa Meem et al.

PDF Code BibTeX

A Point Cloud Transformer for Remote Monitoring and Automated Assessment of Physical Rehabilitation Exercises

IEEE JBHI 2026IF 7.7 · Q1First author

Kazi Rafat Haa Meem, Md. I. Hossain, M. M. L. Elahi, S. Momen, F. Rahman, N. Mohammed, S. Rahman

PDF BibTeX

Mitigating Carbon Footprint in Deep Learning via Knowledge Distillation

PLOS One 2023IF 2.6First author

Kazi Rafat Haa Meem et al.

PDF DOI

DeepMarkerNet: Supervised Duchenne Marker Modeling for Spontaneous Smile Recognition

Pattern Recognition Letters 2024IF 3.9Co-author

Kazi Rafat Haa Meem et al. · Elsevier

PDF DOI

Bangla TTS

Precision-Driven Low-Resource Bangla Text-to-Speech System

ICLR Workshop 2024Co-author

Kazi Rafat Haa Meem et al. · Practical ML for Low-Resource Settings

PDF

🛠️ Projects

MenoChat — Agentic Women's-Health Chatbot

$30K GrantBest Idea & Postericddr,b collaborationLeader

An agentic women's-health assistant (RAG + hybrid LLM generation, STT/TTS, and RL-based dialogue control) for low-literacy users in Bangladesh. Refined with public-health experts at icddr,b icddr,b

, Bangladesh's leading health research institute, so its menstrual and menopausal guidance is medically accurate and culturally sensitive. Best Idea + Best Poster and a $30K grant at NCSRHR 2023.

Point Cloud Transformer for Physical Rehabilitation Assessment

CTRG GrantLeader

Transformer that scores rehabilitation exercise quality from 3D pose and motion data. Funded by a CTRG 2023 research grant (~$4,500).

AdaptBench — Benchmark for LLM Evaluation

Adaption Labs 2026

Selected for the inaugural Adaption Labs grant cohort to build a benchmark for evaluating large language models.

Intelligent Virtual Assistant for Low-Resource Devices

NAACL Mentorship

Multilingual, multimodal voice assistant on constrained hardware using distilled LLMs, with mood recognition and parental controls.

🎤 Talks & Presentations

Interspeech 20262026

Streaming ASR for Code-Switching Speech

Presenting my first-author work on Dynamic Block-Online streaming ASR and the CS-WER metric.

ICLR 2024 · Workshop on Practical ML for Low-Resource Settings2024

Precision-Driven Low-Resource Bangla TTS

Presented our high-fidelity Bangla text-to-speech system with voice cloning at the ICLR workshop.

NCSRHR 20232023

MenoChat — Best Poster Award 🏆

Poster presentation of our agentic women's-health chatbot; won Best Idea + Best Poster.

💼 Experience

Decodis2023 – Jan 2026

Machine Learning Engineer · Boston, MA (Remote)

Designed and shipped multilingual speech and language systems end to end: production ASR for Swahili, Yoruba, Azeri, Bangla, Hindi, and Urdu; real-time transcription with seamless multilingual timestamping; and LLM pipelines for data augmentation and generalization in low-data regimes. Integrated into interview workflows at 1,000-interview scale. Contributed to a pending patent and delivered $45,000+ in cost savings.

Apurba Technologies · NSU R&D Lab2022 – Present

Research Assistant · Dhaka, Bangladesh

Lead researcher on LLMs, ASR, TTS, knowledge distillation, and agentic AI. Directed multilingual data collection and full project lifecycles; mentored junior researchers.

🎓 Education

North South UniversityJan 2019 – May 2022

B.Sc. Computer Science & Engineering · CGPA 3.90/4.00 · Summa Cum Laude

Merit-Based Scholarship throughout (nationwide competitive admission). Focus: Pattern Recognition & Neural Networks, Machine Learning, Probability & Statistics.

🏆 Honors

$30,000 Research Grant + Best Idea + Best Poster — NCSRHR 2023 (MenoChat)
Adaption Labs Inaugural Grant Cohort — 2026 (AdaptBench)
Research featured at Cohere Labs (Cohere For AI) — knowledge-distillation work from PLOS One
CTRG 2023 Research Grant — ~$4,500, Bangladesh national grant
NAACL 2023 Mentorship Program — selected mentee
Summa Cum Laude — North South University (CGPA 3.90/4.00)
Merit-Based Scholarship — North South University (2019)
The Daily Star Award — outstanding O-Levels (2016) & A-Levels (2018)

📎 CV

The full record, on one page.

Download CV →

✉️ Contact

Reach me at kingrafat82@gmail.com. I like meeting people who care about speech, multilingual AI, or health.