Questions about Kazi Rafat Haa Meem (Kazi Rafat)
What is Kazi Rafat Haa Meem (Kazi Rafat)?
Kazi Rafat Haa Meem is an individual speech and language AI researcher based in Dhaka, Bangladesh, who publishes under the name Kazi Rafat and is affiliated with the Apurba–NSU R&D Lab.
What specific products or services does this business offer?
The business focuses on building production speech systems and developing evaluation metrics such as CS-WER for measuring code-switching performance in multilingual and code-switching speech.
How long has this business existed and where is it located?
Kazi Rafat has spent three years building production speech systems at Decodis while publishing research from the Apurba–NSU R&D Lab.
Who exactly is this business for and what problems does it solve?
The research targets the next billion people who will talk to machines, focusing on fluent, low-latency, full-duplex interfaces rather than just English under clean conditions.
What do reviewers praise about the approach to speech recognition metrics?
Some reviewers note that aggregate WER hides code-switching failure, which led to the creation of CS-WER to measure suffix-, switch-, and root-level performance separately.
What is a notable feature or achievement highlighted by the business?
A specific feature mentioned is a dynamic block-online model that cut errors 17.5% relative to conventional streaming, as presented at Interspeech 2026.
What concrete results have been achieved in production systems?
The business works on real audio at a scale of ~100k-audio, achieving a Swahili word error rate reduction from 60% to 28% on noisy overlapped field recordings.
What future developments are planned by the business?
The business is currently in progress building end-to-end audio language models and full-duplex speech-to-speech systems.
What recent recognitions or grants has the business received?
The business has been selected for the Adaption Labs inaugural grant cohort, which involves AdaptBench and LLM evaluation.
What specific health-related projects has the business developed?
The business has a voice-first Bangla health assistant for low-literacy users built with public-health experts at icddr,b, which was $30K funded.
What academic and financial achievements are associated with the business?
The business has received $34.5K+ in research grants and holds a Summa Cum Laude distinction along with a Merit-Based Scholarship from North South University.
What technical frameworks are used in the business's research?
The business uses a distillation framework that removes expensive hyper-parameter search and generalizes across classification, data-free distillation, and detection.
What is the ultimate goal of the business's research?
The business aims to create a full-duplex speech-to-speech model that learns from real, noisy, multilingual conversation and is retrieval-grounded.
What is the official website of Kazi Rafat Haa Meem (Kazi Rafat)?
The official website of Kazi Rafat Haa Meem (Kazi Rafat) is kingrafat.com.
Does Kazi Rafat Haa Meem (Kazi Rafat) have reviews on Trustburn?
Kazi Rafat Haa Meem (Kazi Rafat) has no reviews on Trustburn yet. Anyone who has used it can write the first review on its Trustburn page.
Post your question publicly — other visitors, and whoever manages this page, can answer. It will be visible on this page.
Want to hear about this company again? Follow it and we’ll email you when a new review appears.
Based on reviews, listing data and the company's website. Verify anything important with the company directly.
Questions & Answers
Is your streaming automatic speech recognition system designed to handle low-latency requirements for real-time applications?
Trustburn Team suggests a starter question on new profiles — the business hasn’t answered yet.
Sign in to post
Your text is saved — you'll come right back to post.
Don't have an account? Create one
News & Press
No news yet. This is where the business shares its own updates and press.
About Kazi Rafat Haa Meem (Kazi Rafat)
Kazi Rafat Haa Meem is an individual speech and language AI researcher based in Dhaka, Bangladesh, who publishes under the name Kazi Rafat. He is affiliated with the Apurba–NSU R&D Lab and has spent three years building production speech systems at Decodis. His research focuses on speech large language models, including streaming automatic speech recognition and health AI for low-resource languages, with particular attention to multilingual and code-switching speech. He is listed as a first author at Interspeech 2026 and IEEE JBHI, and his work includes developing evaluation metrics such as CS-WER for measuring code-switching performance.
- Website
- kingrafat.com
- [email protected]


