FSOFT AI Center, FPT Tower, 10 Pham Van Bach, Hanoi (2026)

13/08/2026

📣 FPT AI RESIDENCY RESEARCHER PRESENTS PAPER AT ICME 2026
From July 05 to July 09, 2026, Nguyen Tuan Duc, researcher from the FPT Software AI Center, represented the FPT AI Residency at the IEEE International Conference on Multimedia and Expo (ICME 2026), held in Bangkok, Thailand.
🌟 Established in 2000, ICME is IEEE’s flagship conference in multimedia, jointly sponsored by four IEEE societies: the Circuits and Systems Society, Communications Society, Computer Society, and Signal Processing Society. Ranked A in the ICORE 2026 Conference Rankings, ICME provides an international forum for advances in multimedia technologies, systems, and applications, covering areas such as multimedia analysis, compression and communication, multimodal learning, generative media, immersive systems, and intelligent multimedia applications.
💬 Sharing his experience at the conference, Nguyen Tuan Duc reflected:
“Attending ICME 2026 in Bangkok was an inspiring experience and a valuable opportunity for me to present my research, learn from the international multimedia community, and connect with researchers from both academia and industry.
My strongest impression was that multimedia research is moving beyond traditional content analysis toward systems that can reason, interact, and act in the physical world. This shift was reflected in the strong focus on multimodal reasoning, long-video understanding, embodied AI, world models, trustworthy Vision-Language-Action systems, and generative multimedia technologies.
The three keynote talks also offered complementary perspectives on the future of the field. Professor Lu Fang demonstrated how physical optics and AI can be jointly designed to create more efficient imaging and computing systems. Professor Yonggang Wen emphasized the twin transition of digitalization and sustainability, highlighting both sustainable AI infrastructure and the use of AI to optimize data centers. Professor Antonio Ortega showed how multimedia compression can be optimized for computational metrics and downstream tasks such as detection and segmentation, rather than only for traditional pixel-level quality.
The panel on large multimodal models and Physical AI further reinforced that future AI systems should be evaluated not only by accuracy, but also by robustness, uncertainty, safety, and accountability in real-world environments.
Beyond the technical sessions, I appreciated the opportunity to exchange ideas with researchers from different universities and industrial research labs. Meeting Vietnamese researchers was especially motivating and gave me further encouragement for my own research journey. I am deeply grateful to my supervisor, colleagues, and the FPT Software AI Center for their continued support, which made it possible for me to publish, present, and attend ICME 2026.”
📑 Our team’s research work presented at ICME 2026:
Title: FR-DETR: Frequency and Recurrent Feature Refinement for Robust Object Detection under Adverse Weather
🔗 Preprint: https://arxiv.org/abs/2606.30471
💻 GitHub: https://github.com/Fsoft-AIC/FR-DETR
📊 ICME 2026 by the numbers:
3,989 submissions entered the review process.
1,092 papers were accepted, resulting in an acceptance rate of 27.4%.
The review process involved 12,056 authors from 62 countries and regions, along with 4,680 reviewers, 375 Area Chairs, and 74 Senior Area Chairs
The technical program included 3 keynote talks, 1 panel, 7 tutorials, 10 Grand Challenges, 9 workshops, 26 special sessions, and 2 industrial sessions.

👉 Stay tuned to our fanpage for more updates and research highlights from the FPT AI Residency.

11/08/2026

Millions of lines of code.
Decades of knowledge.
Vietnamese AI is bringing it all back to life. 🇻🇳
At a leading global automotive services company, engineers once spent as much as 40–60% of their time simply searching for information and understanding their software systems.
📍Flezi Metis builds an AI-specific “map” that helps AI understand how the entire system works. As a result, the time engineers spend finding source code has been reduced by around 50%, while task completion time has dropped by around 30%. Following its success in the US, Flezi Metis is expanding into banking, insurance, and the public sector, becoming part of the FleziPT AI ecosystem.
👉 Discover the story of Flezi Metis and how FPT is turning legacy software systems into valuable assets in the comments below.
-------
Hàng triệu dòng code.
Hàng chục năm tri thức.
AI Việt đang đánh thức tất cả. 🇻🇳
Tại một tập đoàn dịch vụ ô tô hàng đầu thế giới, kỹ sư từng dành tới 40–60% thời gian chỉ để tìm thông tin và hiểu hệ thống phần mềm.
📍Flezi Metis xây dựng một “bản đồ” dành riêng cho AI, giúp AI hiểu cách toàn bộ hệ thống vận hành. Kết quả, thời gian đội ngũ kỹ sư tìm mã nguồn giảm khoảng 50%, thời gian hoàn thành công việc giảm khoảng 30%. Sau thành công tại Mỹ, Flezi Metis đang mở rộng sang ngân hàng, bảo hiểm và khu vực công, trở thành một phần của hệ sinh thái AI FleziPT.
👉 Khám phá câu chuyện về Flezi Metis và cách FPT biến những hệ thống phần mềm cũ kỹ thành tài sản vô giá dưới phần bình luận.

20/07/2026

📚 FPT AI RESIDENCY ADVANCES LARGE LANGUAGE MODEL RESEARCH WITH A PAPER PUBLISHED IN TMLR

💫 FPT Software AI Center is delighted to share that a research paper from our AI Residency has been accepted for publication in Transactions on Machine Learning Research (TMLR). This achievement reflects our continued commitment to AI-first research, where residents and mentors work together to advance the foundations of machine learning while making cutting-edge research more accessible to the broader AI community.

👉 Our Published Research Paper
- Title: LibMoE: A Library for Comprehensive Research Mixture of Experts in Large Language Models
- Authors: Nam Nguyen (Batch-5 AI Residency) (1st author), Thong Doan (Batch-5 AI Residency), Luong Tran (Batch-5 AI Residency) , Van Nguyen (mentor), Quang Pham

🔗 GitHub: https://github.com/Fsoft-AIC/LibMoE
https://openreview.net/pdf?id=PB2ju8tq0n

🌐 Transactions on Machine Learning Research (TMLR) is a recognized journal in machine learning, known for its rigorous peer review process and commitment to open scientific exchange. The journal publishes impactful research spanning machine learning theory, algorithms, systems, and real-world applications, providing a trusted platform for advancing the field.

✨ Please follow FPT Software AI Center for more updates on AI-first research and innovations from our AI Residency community.

17/07/2026

🔥 FPT AI RESIDENCY MAKES ITS MARK AT ECCV 2026

💫 FPT Software AI Center is delighted to share that a research paper from our AI Residency community has been officially accepted to the European Conference on Computer Vision (ECCV) 2026. This achievement reflects our continued commitment to AI-first research, where residents and mentors work together to advance fundamental AI while addressing real-world challenges through practical and impactful innovations.

👉 Our Accepted Research Paper
Title: Precise Video-to-Audio Generation with Cross-Modal Alignment in Latent Space
- Authors: Thanh Tran (Batch-4 AI Residency) (1st author), Ngoc-Son Nguyen (Batch-6 AI Residency), Luong Tran (Batch-5 AI Residency), Long-Khanh Pham, Paarth Neekhara, Shehzeen Hussain, Van Nguyen(mentor)
- Preprint: https://arxiv.org/abs/2607.06405
- Code: https://github.com/FSoft-AIC/Flowley

🌐 The European Conference on Computer Vision (ECCV) is one of the world's leading conferences in computer vision, artificial intelligence, and machine learning. Organized by the European Computer Vision Association (ECVA), ECCV brings together researchers and industry leaders to present the latest advances in visual AI, deep learning, and generative models.
For the ECCV 2026 main conference, 10,473 submissions entered the review process, with 2,883 papers accepted for presentation. The 19th European Conference on Computer Vision will take place from September 8–12, 2026 in Malmö, Sweden.
✨ Please follow FPT Software AI Center for more updates on AI-first research and innovations from our AI Residency community.

26/06/2026

💫 FPT AI RESIDENCY ADVANCES ROBOTICS RESEARCH WITH A NEW PAPER ACCEPTED AT IROS 2026

💫 FPT is delighted to share that a research paper from our AI Residency community has been accepted to IROS 2026, one of the world's leading robotics research conferences. This milestone reflects our continued commitment to AI-first research through close collaboration between residents, alumni, and mentors, building strong foundations for practical innovation in robotics and intelligent systems.

👉 Our Accepted Research Paper
- Title: RobotDesign1M: A Large-scale Dataset for Robot Design Understanding
- Authors: Tri Le (Batch-6 AI Residency) (1st author), Toan Nguyen (AI Residency Alumni)*, Quang Tran, Quang Nguyen (AI Residency Alumni)*, Baoru Huang, Hoan Nguyen, Minh Nhat Vu, Tung D. Ta, Anh Nguyen (Mentor)
- Link pre-print: https://arxiv.org/abs/2503.06796
- Link hugging face: https://huggingface.co/datasets/Fsoft-AIC/RobotDesign1M
*Work contributed during time at FPT AI Residency

🌐 The IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) is one of the largest and most impactful robotics research conferences worldwide. Recognized as an A - ranked venue, IROS brings together researchers, engineers, and industry leaders working at the forefront of robotics and intelligent systems.

📍 IROS 2026 will take place in Pittsburgh, USA, from September 27 to October 1, 2026. This year, the conference received 4,348 contributed paper submissions, of which 1,585 papers were accepted, representing an acceptance rate of 36%.

✨ Please follow FPT Software AI Center for more updates on AI-first research and innovations from our AI Residency community.

23/06/2026

🎉 FPT AI RESIDENCY RESEARCHERS PRESENT PAPERS AT CVPR 2026

💡 From June 03 to June 07, 2026, Son Nguyen and Bao Truong, researchers from the FPT Software AI Center, represented the FPT AI Residency at the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026), held in Denver, Colorado, USA.
CVPR is widely recognized as the premier conference in Computer Vision and one of the most prestigious AI conferences globally (rank A*), serving as a leading venue for groundbreaking research in visual understanding, multimodal learning, generative models,...

🌟 Sharing his experience at the conference, Son Nguyen reflected:
“Attending CVPR 2026 in Denver was a memorable experience and a wonderful opportunity to learn and connect with pioneering researchers and companies in AI. Throughout the conference, I had the chance to explore many inspiring works that opened new directions for my future research, spanning audio-visual learning, representation learning, 3D generation, medical imaging, and more. This year, three major trends I witnessed clearly dominated the landscape: 3D, generative models with Diffusion absolutely leading the charge, and vision-language models. What made the experience even more special was seeing not only the global research community, but also Vietnamese professionals from academia and industry making their mark on the world stage, and I feel fortunate to be one of them. I am deeply grateful to my supervisor, my colleagues, and the FPT Software AI Center for their continued support, which made it possible for me to publish at CVPR this year.
The three keynotes at CVPR 2026 offered a broad view of where AI is heading. Dr. Simon Kohl of Latent Labs showed how generative AI is transforming drug discovery by designing molecules from intent rather than screening existing ones. Dr. Jerry Chow from IBM presented how quantum-centric supercomputing, combining quantum hardware, HPC, and AI, is unlocking a computational frontier beyond what any single technology can achieve. Finally, Dr. Thomas Serre challenged the community to look beyond scaling laws, arguing that grounding vision models in biological principles can produce systems that are not only more capable but more aligned with how humans actually see.”

🌟 Bao Truong also reveled his gratitude after attending the conference:
“Attending CVPR 2026 was an inspiring and motivating experience for me, especially because I could directly feel how fast the computer vision community is moving from passive perception toward systems that can reason, interact, and act in the real world.
The keynotes and talks gave me a broader view of this shift: generative AI is becoming a powerful tool for scientific design, future AI systems may rely on stronger computing infrastructure, and artificial vision may need not only larger models but also more human-like and biologically inspired learning principles. In this CVPR 2026, the strongest insight for me was that robotics is becoming a natural convergence point for vision, language, 4D understanding, world models, and action learning. Many discussions suggested that the next generation of vision research will not stop at recognizing objects or reconstructing scenes, but will focus on building embodied agents that can understand physical environments, plan robust actions, and adapt safely to real-world uncertainty.
I also had the privilege of exchanging ideas and connecting with researchers from top-tier universities such as Stanford and CMU, as well as teams from leading companies like NVIDIA Robotics and Meta Reality Labs. In addition, meeting Vietnamese researchers from around the world was genuinely inspiring, as their experiences gave me more motivation for my own research journey. I am very grateful to my mentors, colleagues, and the FPT Software AI Center for their continuous support, which made this valuable experience possible.”

Our team’s research work presented at CVPR 2026:
Title: SIGMA: A Physics-Informed Benchmark for Gas Chimney Understanding in Seismic Images
Authors: Bao Truong (Batch-6 AI Residency) (1st author), Quang Nguyen (AI Residency Alumni), Baoru Huang, Jinpei Han, Van Nguyen (mentor), Ngan Le (mentor), Minh Tan Pham (mentor), Doan Duy Hien (mentor), Anh Nguyen (mentor)
Title: DiFlowDubber: Discrete Flow Matching for Automated Video Dubbing via Cross-Modal Alignment and Synchronization (CVPR Findings)
Authors: Ngoc-Son Nguyen (Batch-6 AI Residency), Thanh Tran (Batch-4 AI Residency), Jeongsoo Choi, Hieu-Nghia Huynh-Nguyen, Truong-Son Hy, Van Nguyen (mentor).

📊 CVPR 2026 by the numbers:
16,092 submissions entered the review process.
4,071 papers were accepted, resulting in an acceptance rate of 25.30%.
The conference welcomed 8,963 in-person attendees and 478 virtual participants from around the world.

📩 Stay tuned to our fanpage for more updates and research highlights from the FPT AI Residency.

15/06/2026

🚀 VOICING THE FUTURE OF AI: FPT AI RESIDENCY CONTINUE STRONG WITH NEW PAPER ACCEPTED AT INTERSPEECH 2026

💫 FPT is proud to share that one research paper from our AI Residency community has been officially accepted to Interspeech 2026 (Long Paper track), one of the world's leading conferences in speech and spoken language technologies.

This achievement reflects the AI-first spirit that drives our researchers to continuously explore new approaches in generative AI and speech intelligence, transforming cutting-edge research into practical solutions for real-world challenges. By advancing both efficiency and quality in AI-generated speech, this work contributes to making next-generation voice technologies more accessible, scalable, and impactful.

👉 Our Accepted Research Paper:

Title: DiFlow-TTS: Compact and Low-Latency Zero-Shot Text-to-Speech with Discrete Flow Matching

Authors: Ngoc-Son Nguyen (Batch-6 AI Residency) (1st author), Thanh Tran (Batch-4 AI Residency), Hieu-Nghia Huynh-Nguyen, Truong-Son Hy, Van Nguyen (mentor)

Abstract: Zero-shot text-to-speech (TTS) has made significant progress in replicating unseen voices, yet balancing generation quality and inference efficiency remains challenging. Autoregressive models suffer from high latency, while diffusion-based approaches are constrained by training-time configurations. Moreover, most flow-based methods operate in continuous space, which introduces optimization challenges because continuous token spaces are inherently more complex than discrete ones. To address this limitation, we propose DiFlow-TTS, a zero-shot TTS framework based on discrete flow matching. The model consists of a deterministic Phoneme-Content Mapper for linguistic modeling and a Factorized Discrete Flow Denoiser that jointly generates prosody and acoustic token streams. Experimental results demonstrate the effectiveness of our approach across multiple evaluation metrics.

💡 Technical Challenge

As AI-generated speech becomes increasingly important across digital content, virtual assistants, and multilingual communication, the demand for voice models that are both high-quality and responsive continues to grow. However, existing zero-shot TTS systems often sacrifice speed for quality, as autoregressive models suffer from high inference latency, while diffusion-based approaches are constrained by training-time configurations. Through DiFlow-TTS, our researchers bring discrete flow matching into the domain of speech generation, establishing it as a compelling new direction for efficient, high-fidelity zero-shot speech synthesis well-suited to resource-constrained and latency-sensitive applications, and highlighting discrete flow models as a promising direction for future generative speech research.

🌐 Interspeech is the world's largest conference dedicated to speech science and spoken language technology, organized by the International Speech Communication Association (ISCA). As a premier venue for speech AI research, the conference brings together leading researchers and practitioners working across speech processing, voice technologies, acoustics, and human-machine communication.

📍 Interspeech 2026 will take place in Sydney, Australia, from September 27 to October 1, 2026, with a strong focus on interdisciplinarity, equity, and linguistic diversity.

✨ Please follow FPT Software AI Center for more updates on AI-first research and innovations from our AI Residency community.

08/06/2026

🔥 FPT's AI TACKLES A HUNDRED-MILLION-DOLLAR CHALLENGE IN OIL&GAS ⛽️🛢️

What does AI look like when it moves beyond the hype and starts solving real-world challenges?

In the oil and gas industry, where every decision carries significant operational and financial implications, our engineers from FPT's AI & Innovation Center (AI1) have developed an AI solution capable of analyzing more than 45 million historical data records to support material evaluation and decision-making.

What once took engineers days, or even weeks, can now be completed in just minutes.

By helping engineers access relevant knowledge faster and navigate thousands of complex variables with greater confidence, AI becomes a powerful assistant rather than a replacement for human expertise.

This is what FPT’s AI-First strategy is all about: applying AI to solve practical business challenges, enhance human capabilities, and create measurable impact for enterprises around the world. From emerging technology to real business value, that is the journey of AI at FPT.

👉Read the full story here: https://en.vneconomy.vn/how-vietnamese-ai-is-reshaping-one-of-oil-and-gass-costliest-challenges.htm

14/05/2026

🔥 FPT AI RESIDENCY OPENS LIMITED RESEARCH POSITIONS FOR A TOP-RANKED AI PROJECT

FPT AI Residency is looking for the next generation of AI researchers to join a newly funded international project:
“Lightweight Multimodal Learning and Human–Robot Interaction”.
The project focuses on developing lightweight AI models for multimodal learning, human–machine/robot interaction, and wearable device optimization. Researchers will work on efficient, real-time AI systems designed for resource-constrained platforms such as wearable devices, edge sensors, mobile systems, and robotic interfaces.

This is a research-oriented opportunity for candidates eager to contribute to novel AI methods, develop prototypes, publish high-quality research, and explore emerging directions in efficient AI, generative models, world models, multimodal learning, and human-centered intelligent systems.

We are now selecting 4 Applied Researchers to work on:
✅ Design and optimize lightweight AI models for deployment on wearable, mobile, robotic, and edge devices.
✅ Develop multimodal learning methods combining data such as vision, audio, motion, physiological signals, tactile data, and contextual sensor inputs.
✅ Build generative AI systems/world models for human–machine interaction and human–robot interaction.
✅ Optimize models for low latency, low power consumption, memory efficiency, and real-time performance.
✅ Develop algorithms for wearable devices, including sensor fusion, signal processing, personalization, adaptive sensing, and on-device inference.
✅ Evaluate AI models in practical human-centered scenarios, including interaction with
machines, robots, assistive systems, or industrial environments.
✅ Collaborate with interdisciplinary teams including international AI researchers, roboticists,
hardware engineers, UX researchers, and domain experts.
✅ Prepare technical reports, research papers, software prototypes, and project documentation, with the goal of publishing high-quality research in top-tier computer vision and robotics venues such as CVPR, ICCV/ECCV, NeurIPS, IROS, and ICRA.

Under the mentorship of Professor Nguyễn Quang Anh from the University of Liverpool, selected researchers will work closely with international researchers and engineers to develop next-generation AI systems designed for real-world deployment.

This opportunity is for candidates who:
📚 Final-year student, Bachelor’s or Master’s degree in Computer Science, Artificial Intelligence, Data Science, or a related field.
📚 Familiarity with machine learning frameworks (e.g., PyTorch, TensorFlow).
📚 Excellent communication skills and a commitment to consistent progress and feedback.
📚 Ability to work independently and collaboratively within a structured research environment.
📚 A growth mindset and curiosity for applying research in real-world contexts.
📚 Prior experience with research, academic writing, or publication (not required but a plus).

Selected researchers will have the chance to:
✨ One-on-one mentorship from top AI researchers and industry experts.
✨ Hands-on experience in research and potential real-world applications.
✨ Opportunity to publish in leading AI venues and build a strong academic profile.
✨ Exposure to AI product development and business innovation processes.
✨ Access to FPT’s internal network of researchers, engineers, and business leaders.
📢 Recruitment Timeline
🗓️ Application Open & Screening: 08 - 22 May 2026
Rolling basis -applications may close early once all positions are filled
🗓️ Projects & Interviews: 18 - 29 May 2026
🗓️ Onboarding: 01 - 07 June 2026
📍 Location: Hanoi & Ho Chi Minh City
📩 Apply now: https://forms.gle/wHKyL7qCyafq2Cp78

FSOFT AI Center

13/08/2026

11/08/2026

20/07/2026

17/07/2026

26/06/2026

23/06/2026

15/06/2026

08/06/2026

14/05/2026

Address

Website

Alerts

Contact The Business

Shortcuts

Share

Category

Listing takedown request