Making AI Work For Everyone: Building A Voice For The Filipino
- Kathleen Co
- May 15
- 4 min read
When Dr. Charibeth Cheng and her colleagues first proposed building a Filipino-English machine translator in 2002, the grant panel’s response was blunt:
“What’s the need for this when people in the Philippines can speak English?”
It was the wrong question. Dr. Cheng knew it then, and two decades of work have proved it. “There might not be a use case for it; there might not be a need for it,” she recalls, “but we still needed to know how to build one, and what the challenges are.”
That answer—patient and quietly defiant—became the foundation of one of the Philippines’ most consequential AI careers. Today, Dr. Cheng is a Full Professor and Associate Dean at De La Salle University’s College of Computer Studies, a co-founder of Senti AI, and the recipient of multiple Google research awards supporting machine translation for Filipino dialects. Her work spans typhoon impact analysis, hate speech detection, and the preservation of the country’s linguistic diversity—Tagalog, Ilocano, Cebuano, Waray, Tausug, and more.
But her most important contribution may be the simplest: she stayed.
A PROBLEM WORTH SOLVING
Dr. Cheng did not set out to be a pioneer. She took a teaching post at DLSU after graduation because she did not want to go into sales, and she discovered she enjoyed it. AI found her, not the other way around.
“DLSU has always had AI in its curriculum,” she explains. “We taught it even during an AI winter, when it seemed like nobody else was teaching it. Few people saw it becoming a reality or having any value.”
That willingness to work on problems without obvious markets defined her early career and ultimately proved prescient. The machine translation project was hard—not just technically, but institutionally. Funding was uncertain. The value of the work was questioned at every turn. Three years of work produced a research outcome, not a commercial product, and half the original team moved on.
But the three who remained built something more durable: expertise that could not be imported. “I wanted to focus on Filipino, on Tagalog, because I felt it was not yet solved,” she says. “All the solutions were English-focused. I wanted to solve this problem.”
"I wanted to focus on Filipino, on Tagalog, because I felt it was not yet solved,” she says. “All the solutions were English-focused. I wanted to solve this problem.”
FILIPINO NLP WAS READY FOR GOOGLE AND BEYOND
When ChatGPT launched and the world suddenly discovered what NLP meant, Dr. Cheng was not scrambling to catch up. She was explaining, patiently, what she had already been doing for nearly two decades.
Most AI systems were trained on vast amounts of English text. They work well for English speakers, but struggle with other languages. For the Philippines, a country with over 180 languages and dialects, this gap is not abstract. When AI tools cannot accurately understand Filipino, Cebuano, or Ilocano, the communities that speak those languages are left behind.
Google eventually noticed. Researchers reached out because the language challenges Dr. Cheng had been quietly solving in Manila were the same ones they were encountering at global scale, and could not resolve from Google headquarters.
"There's actually a large body of work; there's so much Filipino content in the world," she notes, "but it's still not at the level of English or Spanish." The gap exists not because Filipino content is scarce, but because few made it a priority until researchers like Dr. Cheng did.
The broader lesson for project managers extends beyond technology. Every decision we make about suppliers, success metrics, stakeholders, and timelines will have consequences for the people our projects are meant to serve. Dr. Cheng's two decades of work are a reminder that good choices compound quietly. Ask the right questions early, stay committed to the answers, and the returns—though slow—arrive when they matter most.
Senti AI, the company Dr. Cheng co-founded with Ralph Regalado after their Filipino sentiment analyser won Startup Weekend Manila in 2015, was not an easy business to run. They faced a well-resourced multinational competitor. Local enterprise clients were sceptical of a small Philippine firm. The trust that large companies extend to established brands simply was not available to them, regardless of their academic track record.
Senti was acquired by Kollab, a Philippine digital transformation company, in December 2024. Dr. Cheng frames it not as a retreat, but as a logical conclusion: a Filipino AI product built for Philippine industries, finding its home in a Philippine company. The principle held even in the exit.
"We really wanted it to be absorbed by a Philippine company,” she says, “kasi our services are really for the Philippines.”
For Filipino project managers watching colleagues leave for better-paying roles abroad, this is perhaps the most grounded inspiration Dr. Cheng offers. She is not naïve about structural barriers: bureaucracy, underfunding, and the slow pace at which academic research reaches the public. She names them plainly. But she also demonstrates, through twenty years of choices, that staying and building is not martyrdom; it is strategy.
The Question Worth Asking
The DOST grant panel in 2002 asked whether there was a need. Dr. Cheng answered with a different question: if no one built this, would anyone in the Philippines ever know how?
That shift in thinking—from “is there demand today” to “what are we building for tomorrow”—is the mindset that separates reactive project delivery from genuinely sustainable technology leadership.
For project managers navigating AI adoption in their own organisations, the lesson is transferable. The systems being built today carry assumptions about language, about users, and about whose needs count. Those assumptions have consequences. Examining them before the build, not after, is not idealism—it is good project management.
Dr. Cheng says she was in the right place at the right time. But place and time, it turns out, are things you can build towards.
