FSOFT AI Center

23/03/2026

🔥 FPT AI RESIDENCY EXPANDS ROBUST VISION RESEARCH WITH A PAPER AT ICME 2026

💫 FPT is delighted to share that a research paper from our AI Residency community has been accepted to the IEEE International Conference on Multimedia & Expo (ICME) 2026, a leading venue for multimedia technologies, systems, and applications. This achievement reflects the continued efforts of the AI Residency program in cultivating AI-first research, where fundamental advances in machine learning and computer vision are developed with a strong focus on real-world impact.

👉 Our Accepted Research Paper
Title: FR-DETR: Frequency and Recurrent Feature Refinement for Robust Object Detection under Adverse Weather

Authors:
Tuan-Duc Nguyen (Batch-5 AI Residency), Duc-Trong Le (mentor)

Abstract:
Object detection under adverse weather remains challenging due to severe visual degradations and domain shifts. Existing enhancer-based approaches attempt to improve detection by cascading an enhancer with a detector, but they introduce redundant feature extraction and incur high computational cost with limited accuracy gains when pair with SOTA detectors.
We propose FR-DETR, a detector-centric framework that refines features rather than images, focusing enhancement on regions of interest and leveraging frequency-domain cues. Specifically, we design (I) a Frequency Refinement Module that dynamically separates and reweights low- and high-frequency components to improve foreground-background discrimination, and (II) a Recurrent Focus Refinement Module (RFRM) that iteratively refines features using coarse predictions as guidance.
Extensive experiments demonstrate that FR-DETR achieves superior detection accuracy under adverse weather while being significantly more computationally efficient than enhancer-based methods.

🌐 The IEEE International Conference on Multimedia & Expo (ICME) is a flagship multimedia conference jointly sponsored by four IEEE societies: Circuits and Systems (CAS), Communications (ComSoc), Computer (CS), and Signal Processing (SPS). Since its launch in 2000, ICME has served as an important forum for presenting advances in multimedia technologies, systems, and applications across research communities.
This 2026, ICME received 3,810 valid submissions and accepted 1,101 papers, resulting in an acceptance rate of only 28.89%.
📍ICME 2026 will take place from July 5-9, 2026 in Bangkok, Thailand.

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

18/03/2026

🔥 FPT AI RESIDENCY ADVANCES MULTIMODAL AI RESEARCH WITH A PAPER ACCEPTED AT CVPR 2026

💫 FPT is enchanted to share that one research paper from our AI Residency community has been accepted to the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2026 Findings. This work contributes to advancing AI-driven multimedia technologies, particularly in automated video dubbing where generating expressive speech while maintaining precise synchronization with visual cues remains a challenging task.

👉 Our Accepted Research Paper
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).

Abstract:
Video dubbing has broad applications in filmmaking, multimedia creation, and assistive speech technology. Existing approaches either train directly on limited dubbing datasets or adopt a two-stage pipeline that adapts pre-trained text-to-speech (TTS) models, which often struggle to produce expressive prosody, rich acoustic characteristics, and precise synchronization. To address these issues, we propose DiFlowDubber with a novel two-stage training framework that effectively transfers knowledge from a pre-trained TTS model to video-driven dubbing, with a discrete flow matching generative backbone. Specifically, we design a FaPro module that captures global prosody and stylistic cues from facial expressions and leverages this information to guide the modeling of subsequent speech attributes. To ensure precise speech-lip synchronization, we introduce a Synchronizer module that bridges the modality gap among text, video, and speech, thereby improving cross-modal alignment and generating speech that is temporally synchronized with lip movements. Experiments on two primary benchmark datasets demonstrate that DiFlowDubber outperforms previous methods across multiple metrics.

Demo Page: https://nngocson2002.github.io/projects/diflowdubber

🌐 CVPR is widely recognized as one of the most influential conferences in computer vision and pattern recognition, bringing together researchers from academia and industry around the world. The conference features advances across a wide spectrum of topics, from visual understanding and generative modeling to multimodal learning and practical AI applications. In 2026, 16,092 submissions entered the review process, with 4,090 papers ultimately accepted, resulting in an acceptance rate of 25.42%.
📍CVPR 2026 will be held from June 3-7, 2026 in Denver, Colorado, USA

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

16/03/2026

👏 FPT SOFTWARE AI CENTER ACHIEVED DUAL HONORS AT PROCESS DAY 2026: QUALITY AMBASSADOR & PROCESS AMBASSADOR!
At FPT Software’s annual Process Day on March 16, FPT Software AI Center proudly celebrated a major milestone as Mr. Huỳnh Tấn Dưỡng and Mr. Hoàng Văn Trưởng were honored with the titles Quality Ambassador of the Year and Process Ambassador of the Year.

Through their dedicated efforts to enhance quality standards and optimize the company’s process framework, both have contributed significantly to improving project efficiency while advancing the spirit of Quality & Process Excellence—a core value that has shaped FPT Software for over two decades.
-------
1️⃣ Mr. Huỳnh Tấn Dưỡng - “BÙI QUANG NGỌC – QUALITY AMBASSADOR” Award
The “Bùi Quang Ngọc – Quality Ambassador” Award recognizes Mr. Dưỡng’s outstanding contributions to quality governance in the Fixed Price project, helping enhance deliverable quality, minimize risks, and increase customer satisfaction.

Serving as the Project Manager, he has been leading the development of an AI model designed for weather forecasting in the Tokyo area, with a particular focus on solar radiation prediction - a critical dataset that supports communities in better preparing for and responding to climate change.

The project involves several complex components, including:
- Collecting simulation data from the client and conducting data labeling for AI model training
- Developing a Surrogate AI Model based on the Stable Diffusion architecture, incorporating three AI models
- Building a user-friendly interface that allows users to input physical parameters for forecasting purposes

By ensuring rigorous quality standards throughout the entire project lifecycle, Mr. Dưỡng has played a key role in enhancing the reliability of the AI product while optimizing project implementation effectiveness.
-------
2️⃣ Mr. Hoàng Văn Trưởng – “HÙNG RÂU – PROCESS AMBASSADOR” Award
The “Hùng Râu – Process Ambassador” Award recognizes Mr. Trưởng’s exceptional contributions to improving and effectively implementing process systems at the project, unit, and company-wide levels.

Currently serving as an AI Scientist at FPT Software AI Center, Mr. Trưởng has played an important role in strengthening FPT Software’s quality management system, particularly through his contributions to achieving ISO/IEC 42001, the international standard for AI Management Systems (AIMS).

One of his most notable achievements is the development of the AIMS Handbook, a comprehensive framework consisting of 72 documents, including:
- Processes
- Templates
- Guidelines

This documentation set has become a critical foundation for standardizing the management and development of AI systems across FSOFT, contributing to improved operational efficiency and supporting sustainable AI development within the organization.
-------
FPT Software Process Day, held annually on March 16, commemorates a significant milestone in 2002, when FPT Software successfully achieved CMM Level 4 certification. At that time, the company became the first enterprise in Southeast Asia to reach this level and was recognized among the top 100 companies worldwide for quality management.

Since then, Process Day has been organized every year to promote a strong process-driven culture, honor individuals and teams who contribute to software quality, and reaffirm FPT Software’s position in the global technology market.
-------
👏 Congratulations once again to Mr. Huỳnh Tấn Dưỡng and Mr. Hoàng Văn Trưởng - two outstanding ambassadors from AI Center who continue to carry forward FPT Software’s journey of Quality & Process Excellence.

11/03/2026

🔥 FPT DevOps Engineer Joins the World’s Elite Golden Kubestronauts, Among Only 5 Vietnamese to Ever Achieve this Level

💫 FPT is proud to announce that Nguyen Huynh Duc, a DevOps Engineer from the AI4SE (AI for Software Engineering) team at FPT AI Center, has officially earned the prestigious Golden Kubestronaut title - the highest distinction in the global Kubernetes and Cloud Native certification ecosystem granted by the Cloud Native Computing Foundation (CNCF).

This milestone places Duc among an exceptionally small group of global experts and makes him one of only five Vietnamese engineers ever to achieve this recognition. He is also the first engineer from FPT Software to reach the Golden Kubestronaut level, marking a significant milestone for FPT’s engineering community. Duc’s achievement highlights not only his individual dedication and technical excellence but also stands as a powerful example of how FPT engineers are increasingly contributing to and being recognized within the world’s most advanced technology ecosystems.

🌐 The Kubestronaut program, operated by the Cloud Native Computing Foundation under the Linux Foundation, recognizes engineers who demonstrate deep, comprehensive mastery across the Cloud Native technology ecosystem. As the organization stewarding Kubernetes and more than 200 of the world’s most influential Cloud Native open-source projects, CNCF created this program to honor professionals who commit to continuous, high-level technical advancement.

✨ Introduced in April 2025 at KubeCon + CloudNativeCon Europe, the Golden Kubestronaut designation represents the highest tier of expertise within the program. Globally, 3,359 engineers have been recognized as Kubestronauts, yet only 384 have reached the Golden tier — meaning fewer than 1/9 have achieved this elite status.

🌐 Even earning the Kubestronaut title itself already requires passing five core Kubernetes certifications, a challenge that demands extensive hands-on experience and deep architectural understanding. Reaching the Golden Kubestronaut level is an extraordinary achievement. Candidates must successfully pass all 16 CNCF Cloud Native certifications, covering the full spectrum of modern cloud-native engineering: Kubernetes administration, application development, security, observability, GitOps, service mesh, networking, policy governance, platform engineering, and Linux system administration, along with the Linux Foundation Certified Sysadmin (LFCS) credential. All certifications must remain valid at the same time, a requirement that reflects sustained mastery rather than a one-time accomplishment.

In the development of AI, cloud-native technologies play the role of a “backbone” that enables the transition of AI models from the research stage to real-world production systems capable of operating reliably and serving millions of users. Engineers with deep expertise in Kubernetes and the Cloud Native ecosystem are essential to turning AI innovations into practical products. They possess the capabilities not only to design and operate large-scale infrastructure, but also to optimize deployment, scalability, and system management. They help bridge the gap between AI research and real-world applications, ensuring that research breakthroughs can be quickly transformed into stable, scalable, and efficient products.

--------

🔥 Kỹ sư DevOps của FPT gia nhập nhóm “Golden Kubestronaut” tinh hoa toàn cầu – trở thành một trong 5 người Việt đầu tiên đạt danh hiệu này

FPT chính thức chinh phục một cột mốc đáng tự hào khi Nguyễn Huỳnh Đức, DevOps Engineer thuộc nhóm AI4SE (AI for Software Engineering) tại FPT AI Center, chính thức đạt danh hiệu Golden Kubestronaut - danh hiệu cao nhất trong hệ thống chứng chỉ Kubernetes và Cloud Native toàn cầu, được cấp bởi Cloud Native Computing Foundation (CNCF).

Thành tích này đã đưa Đức vào nhóm chuyên gia công nghệ hàng đầu thế giới và trở thành một trong 5 kỹ sư Việt Nam xuất sắc đạt danh hiệu Golden Kubestronaut. Đồng thời, anh cũng là kỹ sư đầu tiên của FPT Software đạt được cấp độ này. Thành tựu của Nguyễn Huỳnh Đức không chỉ thể hiện sự nỗ lực cá nhân và năng lực kỹ thuật hàng đầu, mà còn là minh chứng cho việc các kỹ sư FPT ngày càng khẳng định vị thế và đóng góp vào những hệ sinh thái công nghệ tiên tiến nhất trên thế giới.

🌐 Chương trình Kubestronaut do CNCF, tổ chức trực thuộc Linux Foundation, vận hành nhằm vinh danh các kỹ sư có năng lực chuyên sâu và toàn diện trong hệ sinh thái công nghệ Cloud Native. CNCF hiện là đơn vị quản lý Kubernetes cùng hơn 200 dự án mã nguồn mở có ảnh hưởng lớn trong lĩnh vực cloud-native. Chương trình này được CNCF thiết kế để ghi nhận những cá nhân không ngừng nâng cao trình độ kỹ thuật ở mức cao nhất.

✨ Danh hiệu Golden Kubestronaut lần đầu được giới thiệu vào tháng 4/2025 tại sự kiện KubeCon + CloudNativeCon Europe và trở thành đại diện cho cấp độ chuyên môn cao nhất trong chương trình. Trên toàn thế giới hiện có 3.359 kỹ sư đạt danh hiệu Kubestronaut, nhưng chỉ 384 người đạt đến cấp độ Golden, điều đó đồng nghĩa với chỉ khoảng 1/9 số Kubestronaut có thể chạm tới cấp độ này.

Việc đạt danh hiệu Kubestronaut cũng đã là một thử thách lớn với các kỹ sư, khi ứng viên phải vượt qua 5 chứng chỉ Kubernetes cốt lõi, đòi hỏi kinh nghiệm thực tế sâu rộng và hiểu biết kiến trúc hệ thống vững chắc. Sau đó, để có thể trở thành Golden Kubestronaut, ứng viên buộc phải hoàn thành toàn bộ 16 chứng chỉ Cloud Native của CNCF, bao phủ hầu hết các lĩnh vực quan trọng của kỹ thuật cloud-native hiện đại như: quản trị Kubernetes, phát triển ứng dụng cloud-native, bảo mật, quan sát hệ thống (observability), GitOps, service mesh, networking, policy governance, platform engineering và quản trị hệ thống Linux. Quan trọng hơn hết, tất cả các chứng chỉ này phải đồng thời còn hiệu lực. Điều này phản ánh năng lực chuyên môn được duy trì liên tục của các kỹ sư chứ không chỉ là thành tích đạt được trong một lần.

🌐 Trong quá trình phát triển AI, Cloud-native đóng vai trò như “xương sống” giúp chuyển đổi một mô hình AI từ giai đoạn nghiên cứu (research) sang sản phẩm thực tế (production) có thể vận hành ổn định và phục vụ hàng triệu người dùng. Vì vậy, những kỹ sư có kiến thức chuyên sâu về Kubernetes và hệ sinh thái Cloud Native thường có năng lực cao trong việc hiện thực hóa các mô hình AI thành sản phẩm. Họ không chỉ hiểu cách xây dựng và vận hành hạ tầng quy mô lớn, mà còn có khả năng tối ưu quá trình triển khai, mở rộng và quản lý hệ thống, giúp các kết quả nghiên cứu nhanh chóng trở thành những ứng dụng thực tiễn, ổn định và hiệu quả.

😍 FPT DevOps engineer joins the global elite “Golden Kubestronaut”

Nguyen Huynh Duc, a DevOps engineer at FPT Software AI Center, FPT Corporation, has recently become one of the first five Vietnamese to earn the Golden Kubestronaut title – the highest level in the Kubernetes certification system recognized by the Cloud Native Computing Foundation (CNCF). To date, only 384 engineers worldwide have reached this distinction.

✨ The achievement further affirms the expertise of FPT’s engineering team in the cloud-native domain – a core technological foundation that enables AI systems to transition from research to real-world applications serving millions of users. With a team of AI engineers well-versed in cloud-native technologies and actively contributing to the development of advanced technology ecosystems worldwide, FPT is steadily expanding its capabilities in research and mastery of next-generation strategic technologies.
--------

😍 Kỹ sư DevOps FPT gia nhập nhóm tinh hoa toàn cầu “Golden Kubestronaut”

Mới đây, Nguyễn Huỳnh Đức – kỹ sư DevOps tại FPT Software AI Center, Tập đoàn FPT – vừa trở thành một trong 5 người Việt Nam đầu tiên đạt danh hiệu Golden Kubestronaut, cấp độ cao nhất trong hệ thống chứng chỉ Kubernetes do Cloud Native Computing Foundation (CNCF) chứng nhận. Trên toàn thế giới hiện chỉ có 384 kỹ sư đạt tới cấp độ này.

✨ Thành tích trên tiếp tục khẳng định năng lực chuyên môn của đội ngũ kỹ sư FPT trong lĩnh vực cloud-native – nền tảng công nghệ lõi giúp chuyển đổi các hệ thống AI từ giai đoạn nghiên cứu sang ứng dụng thực tế phục vụ hàng triệu người dùng. Với đội ngũ kỹ sư AI am hiểu cloud-native và đang tham gia phát triển các hệ sinh thái công nghệ tiên tiến trên thế giới, FPT đang từng bước mở rộng năng lực nghiên cứu và làm chủ các công nghệ chiến lược thế hệ mới.

05/03/2026

🔥 ACM SIGSOFT DISTINGUISHED PAPER AWARD RECOGNIZES FPT AI RESIDENCY RESEARCH AT ICSE 2026

💫 We are proud to share that our AI Residency community has achieved a landmark success: our research paper has been accepted to the International Conference on Software Engineering (ICSE 2026) and honored with the ACM SIGSOFT Distinguished Paper Award. This work was selected among the Top 1.5% of submissions (22 out of 1469), underscoring its technical excellence, clarity, and transformative potential.

👉 Our Awarded Research Paper
Title: SWE-Synth: Synthetic Repo-level Bug Dataset for Training Automated Program Repair Models

Authors: Minh Pham* (AI Residency Alumni), Huy Phan* (AI Residency Alumni), Hoang Phan, Cuong Le (AI Residency Alumni), Tien N. Nguyen (mentor), Nghi Bui (mentor)
*co-first-author

Abstract: Large language models (LLMs) are revolutionizing automated program repair (APR) by localizing bugs, generating patches, and verifying fixes. Yet progress has been constrained by the absence of scalable, verifiable datasets. SWE-Synth introduces a groundbreaking framework for synthesizing realistic, repository-level bug-fix datasets enriched with test cases and structured repair trajectories. By leveraging LLM agents to simulate debugging workflows, SWE-Synth generates high-quality, process-aware data at scale - a breakthrough that strengthens the real-world viability of autonomous bug-fixing systems and accelerates innovation in software engineering automation.

🌐 ICSE - rank A* is widely recognized as the premier global forum for software engineering*, bringing together top-tier researchers, practitioners, and educators to share cutting-edge advancements, trends, and real-world experiences. In 2026, ICSE will take place from April 12 - 18 in Rio de Janeiro, Brazil, with the core conference running April 15 -17.

✨ This achievement not only underscores the strength of our AI‑first research but also demonstrates how synthetic, agent‑generated datasets can fundamentally reshape the future of autonomous software repair, driving greater reliability, productivity, and innovation across the industry. The influence of SWE‑Synth reaches far beyond its technical contributions: by offering a scalable benchmark, it accelerates progress in AI‑driven software reliability and strengthens the development of autonomous bug‑fixing models. At the same time, the dataset empowers the creation of AI‑powered developer tools capable of intelligently debugging, patching, and verifying code, reducing manual effort while boosting efficiency. For FPT Software and its clients, integrating SWE‑Synth into AI‑first workflows paves the way for next‑generation engineering solutions that enhance reliability, shorten development cycles, and lower costs. Beyond enterprise impact, its open‑source availability catalyzes broader innovation, enabling the global research community to push forward the frontier of AI‑powered software development.

24/02/2026

🔥 FPT AI RESIDENCY ADVANCES SEISMIC RESEARCH WITH A PAPER ACCEPTED AT CVPR 2026

💫 FPT is delighted to share that one research paper from our AI Residency community has been officially accepted to the IEEE CVF Conference on Computer Vision and Pattern Recognition CVPR 2026. This work marks an important step in applying AI-first research to real-world challenges in the Oil and Gas domain, where reliable interpretation directly impacts safety and operational decisions.

👉 Our Accepted Research Paper
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)

Abstract:
Seismic images reconstruct subsurface reflectivity from field recordings, guiding exploration and reservoir monitoring. Gas chimneys are vertical anomalies caused by subsurface fluid migration. Understanding these phenomena is crucial for assessing hydrocarbon potential and avoiding drilling hazards. However, accurate detection is challenging due to strong seismic attenuation and scattering. Traditional physics-based methods are computationally expensive and sensitive to model errors, while deep learning offers efficient alternatives, yet lacks labeled datasets. In this work, we introduce \textbf{SIGMA}, a new physics-based dataset for gas chimney understanding in seismic images, featuring (i) pixel-level gas-chimney mask for detection and (ii) paired degraded and ground-truth image for enhancement. We employed physics-based methods that cover a wide range of geological settings and data acquisition conditions. Comprehensive experiments demonstrate that SIGMA serves as a challenging benchmark for gas chimney interpretation and benefits general seismic understanding.

🌐 CVPR is one of the leading venues in computer vision and pattern recognition, gathering researchers from academia and industry worldwide. The conference covers a broad range of topics, including visual understanding, generative models, multimodal learning, and real-world AI applications. In 2026, 16,092 submissions were reviewed, and 4,090 papers were accepted, resulting in an overall acceptance rate of only 25.42%.
📍CVPR 2026 will be held from June 3-7, 2026 in Denver, Colorado, USA

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

13/02/2026

Kicking off a new chapter with pride and purpose 🚀

From Research breakthroughs to world-class Talent, impactful Initiatives, and sustainable Business growth — every milestone reflects our AI-first DNA.

We’re building, scaling, and innovating with one mission in mind: to lead the AI disruption in Vietnam 🇻🇳🤖

🎬 Hit play to relive a memorable and outstanding 2025 - a year of bold moves, big wins, and meaningful impact - and moving forward to 2026 - a year of AI disruption from FPT.

As we step into the 2026 Lunar New Year, we’re grateful for the journey so far and excited for what’s ahead.

✨ Chúc Mừng Năm Mới — may this year bring bold ideas, breakthrough innovation, and shared success for us all.

09/02/2026

🎉 FPT SOFTWARE AI RESIDENCY CONTINUES STRONG WITH A PAPER ACCEPTED AT ICRA 2026

💫 Starting 2026 with another encouraging research milestone, FPT Software AI Center is delighted to share that one research paper by our AI Residency members and mentors has been officially accepted to the IEEE International Conference on Robotics and Automation ICRA 2026 (rank A*).
This work reflects our continued focus on building AI-first foundations for robotics that are efficient interpretable and grounded in real-world applications.

👉 Our Accepted Research Paper
Title: SlotVLA: Towards Modeling of Object-Relation Representations in Robotic Manipulation

Authors (* co-first author, † corresponding author):
Taisei Hanyu*, Nhat Chung†,* (AI Residency Alumni), Huy Le (AI Residency Alumni), Toan Nguyen (AI Residency Alumni), Yuki Ikebe, Anthony Gunderman, Duy Nguyen Ho Minh, Khoa Vo, Tung Kieu (mentor), Kashu Yamazaki, Chase Rainwater, Anh Nguyen (mentor), Ngan Le (mentor)

Abstract:
Inspired by how humans reason over discrete objects and their relationships, we explore whether compact object-centric and object-relation representations can form a foundation for multitask robotic manipulation. Most existing robotic multitask models rely on dense embeddings that entangle both object and background cues, raising concerns about both efficiency and interpretability. In contrast, we study object-relation-centric representations as a pathway to more structured, efficient, and explainable visuomotor control. Our contributions are two-fold. First, we introduce LIBERO+, a fine-grained benchmark dataset designed to enable and evaluate object-relation reasoning in robotic manipulation. Unlike prior datasets, LIBERO+ provides object-centric annotations that enrich demonstrations with boxand mask-level labels as well as instance-level temporal tracking, supporting compact and interpretable visuomotor representations. Second, we propose SlotVLA, a slot-attention-based framework that captures both objects and their relations for action decoding. It uses a slot-based visual tokenizer to maintain consistent temporal object representations, a relation-centric decoder to produce task-relevant embeddings, and an LLM-driven module that translates these embeddings into executable actions. Experiments on LIBERO+ demonstrate that object-centric slot and object-relation slot representations drastically reduce the number of required visual tokens, while providing competitive generalization. Together, LIBERO+ and SlotVLA provide a compact, interpretable, and effective foundation for advancing object-relation-centric robotic manipulation.

🌐 ICRA is the flagship conference of the IEEE Robotics and Automation Society and one of the most prestigious venues in robotics worldwide. The conference attracts a diverse community of researchers, engineers, and industry leaders, with a strong focus on advancing autonomous systems, intelligent robotics, and their practical applications across industries.
📍This year, ICRA 2026 will take place from June 1–5 2026 in Vienna, Austria

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

06/02/2026

🎉 FPT SOFTWARE AI RESIDENCY KICK OFF 2026 WITH 6 PAPERS ACCEPTED AT ICLR 2026

💫 Kicking off 2026 with encouraging research milestones, FPT Software AI Center is delighted to share that 6 research papers by our AI Residency members, mentors and alumni have been officially accepted to the International Conference on Learning Representations (ICLR) 2026 - the premier gathering of professionals dedicated to the advancement of the branch of artificial intelligence.
These works reflect the AI1 spirit of our program where fundamental research, scalable systems and real-world relevance go hand in hand.

👉 Our Accepted Research Papers (Full abstracts are available in the comment):
1. Quasi-Equivariant Metanetworks
Author: Viet-Hoang Tran*, An Nguyen The* (Batch-5 AI Residency), Benoît Guérand, Thieu Vo (mentor), Tan Minh Nguyen
*: co-first authors
2. CLIP-FMoE: Scalable CLIP via Fused Mixture-of-Experts with Enforced Specialization
Authors: Luong Tran (Batch-6 AI Residency) (first author), Lan-Cuong Nguyen (Batch-5 AI Residency), Huynh Dang Nguyen (Batch-6 AI Residency), Dat Nguyen Cong (Batch-5 AI Residency), Dung Le (mentor) , Van Nguyen (mentor)
3. Mixed-Curvature Tree-Sliced Wasserstein Distance
Authors: Duy-Tung Pham* (Batch-5 AI Residency), Viet-Hoang Tran*, Thieu Vo (mentor), Tan Nguyen
*: co-first author
4. Attention Is All You Need for KV Cache in Diffusion LLMs
Authors: Quan Nguyen-Tri* (AI Residency Alumni), Mukul Ranjan*, Zhiqiang Shen
*: co-first authors
5. Tree-sliced Sobolev IPM
Authors: Viet-Hoang Tran*, Thanh Tran*, Thanh Chu, Duy-Tung Pham (Batch-5 AI Residency), Trung Khang Tran, Tam Le**, Tan Nguyen**
6. Revisiting Tree-Sliced Wasserstein Distance Through the Lens of the Fermat–Weber Problem
Authors: Viet-Hoang Tran*, Thanh Tran*, Thanh Chu*, Trung-Khang Tran, Duy-Tung Pham(Batch-5 AI Residency), Tam Le**, Tan Nguyen**

🌐 ICLR is globally renowned as the most influential conferences in deep learning and representation learning. ICLR 2026 received around 19,000 submissions with an overall acceptance rate of approximately 28%.
📍This year, the 14th conference will take place from April 23–27 2026 in Rio de Janeiro, Brazil.

✨ Follow us for more updates on our journey in AI and technology research!

30/01/2026

🌟 AI-first Impact: Turning Innovation into Real Value 🌟

In 2025, AI at FPT was no longer approached as a standalone technology, but as a way of working embedded across teams, solutions, and decision making. An AI-first mindset gradually reshaped how ideas moved from concept to deployment, how collaboration happened across functions, and how value was created in real business contexts.

Rather than focusing solely on innovation itself, the emphasis shifted toward applying AI where it could make a meaningful difference, strengthening long-term partnerships and delivering outcomes that extend beyond individual projects. This evolution reflects a more mature stage of AI adoption, where impact is defined by consistency, trust, and real-world relevance.

As FPT moves into 2026, this journey continues with a sharper focus on scaling impact and building human-centric AI that delivers value where it matters most.

👉 Explore the infographic to see how AI-first impact took shape in 2025.