Agha Ali Raza

Speech & Language Technologies · AI for the under-served

Agha Ali Raza

Tenured Associate Professor of Computer Science, LUMS · Founding Director, CSaLT · Speech & Language Vertical Lead, National AI Hub

I build speech and language AI for the people technology usually leaves behind: communities that are low-literate, low-income, or linguistically marginalized. My systems run over ordinary phone calls, with no internet or smartphone required, and have reached more than 300,000 people across Pakistan, India, and West Africa. My goal is a digital world that is accessible, inclusive, equitable, and safe for everyone.

Agha Ali Raza
300,000+
people reached
71
publications
USD 5.8M+
research funding
3,000+
professionals trained on AI

What I work on now

National AI Hub · Gates Foundation

Production speech & language for Pakistan's languages

As Speech & Language Technologies Vertical Lead, I am building production-grade speech recognition, text-to-speech, machine translation, and medically contextualized large language models for Urdu, Punjabi, Sindhi, Balochi, and Pashto, all aimed at maternal, newborn, and child health.

Audio security & forensics

Deepfake defense and speech forensics

Through the Crime Investigation and Prevention Lab (CIPL), I develop explainable, data-efficient audio deepfake detection and core Urdu speech forensics (recognition, keyword spotting, speaker ID) to counter AI-driven fraud and protect tech-naive users.

Generative AI literacy

Demystifying AI for 3,000+ professionals

Since 2023, more than 85 training programs for over 3,000 people, from C-suite leaders, vice-chancellors, and 300 doctors to faculty, civil servants, and students, on using generative AI productively and reckoning honestly with its limits and ethics.

The through-line

The Internet of the Orals

Roughly half the world is still offline. My work grounds voice-first telephone services in entertainment and conversation so that they spread on their own through under-connected communities, carrying health information, job opportunities, and a social voice to people who can neither read nor afford a smartphone. Seeded with a handful of users, systems like Polly, Super Abbu, and Baang have grown virally to reach over 300,000 people across Pakistan, India, and West Africa.

Read about the research →   See the deployed systems →

Recent highlights

2026
Appointed Speech & Language Technologies Vertical Lead of the National AI Hub for MNCH, a Gates Foundation-funded national initiative (USD 4.2M).
2026
New papers at Interspeech 2026 and EACL 2026, and in Scientific Reports (Nature Portfolio) and the Journal of Development Economics. All publications →
2025
PakBBQ (EMNLP 2025), plus an active-learning line of work on efficient machine translation and speech.
2024
Tenured and promoted to Associate Professor; won the FAST-NU Alumni Achievement Award; served as Program Committee co-chair for ACM COMPASS 2024.
2023–24
Received the LUMS Vice Chancellor's Award for Teaching Excellence.
LUMS CSaLT Carnegie Mellon University