top of page
LATEST TECH ARTICLES
Galaxy A-Series Gemma3 Pipeline Benchmark
Why This Test Matters One SoC generation changed inference speed by 29%. We ran Gemma3 270M INT8 on four Galaxy A-series devices to find out where on-device LLM becomes practically usable. We tested gemma-3-270m-it-int8 via MediaPipe CPU backend on the Galaxy A16, A26, A36, and A56, measuring latency, token throughput, memory, and accuracy across 25 prompts. We also compared parallel (all 4 devices simultaneously) vs serial (each device independently, 2 runs) execution to ver
Apr 14
Gemma 3n vs Gemma 4: A Real-World Benchmark Guide on the Galaxy S25 Ultra
Click the image to view the report. Every time Google ships a new generation of the Gemma series, the question we ask first is simple: how much faster does it actually run on real hardware? To answer that directly, TecAce ran a head-to-head benchmark between Gemma 3n and Gemma 4 on the Samsung Galaxy S25 Ultra under identical conditions. We tested four model configurations using the llama.cpp CPU inference engine: the previous-generation Gemma 3n E2B Q8_0 as our baseline, a
Apr 14
Galaxy S25 vs S26: On-Device AI Performance Benchmark
Galaxy S25 vs S26: On-Device AI Performance Benchmark Reversal! (Snapdragon 8 Elite Gen 1 vs Gen 2) Does a newer chipset always guarantee faster AI performance? Based on real-world test data conducted by TecAce , we compared the on-device LLM performance between the Galaxy S25 and the Galaxy S26 to find out. Test Overview Devices Compared : Galaxy S25 (Snapdragon 8 Elite) vs. Galaxy S26 (Snapdragon 8 Elite Gen 2) Test Models : Gemma3 1B (INT4) : An ultra-lightweight conversat
Mar 30
![[On-Device AI Chatbot] Part 10: The Future of On-Device AI and TecAce's Roadmap (Conclusion)](https://static.wixstatic.com/media/2ea07e_d1771a9889764093a8c855756693ba51~mv2.png/v1/fill/w_745,h_559,fp_0.50_0.50,q_35,blur_30,enc_avif,quality_auto/2ea07e_d1771a9889764093a8c855756693ba51~mv2.png)
![[On-Device AI Chatbot] Part 10: The Future of On-Device AI and TecAce's Roadmap (Conclusion)](https://static.wixstatic.com/media/2ea07e_d1771a9889764093a8c855756693ba51~mv2.png/v1/fill/w_745,h_559,fp_0.50_0.50,q_95,enc_avif,quality_auto/2ea07e_d1771a9889764093a8c855756693ba51~mv2.png)
[On-Device AI Chatbot] Part 10: The Future of On-Device AI and TecAce's Roadmap (Conclusion)
The Future of On-Device AI and TecAce's Roadmap Throughout this 9-part series, we have chronicled the entire journey of developing an on-device chatbot—a solution to cloud cost and data security issues. We covered everything from selecting a Small Language Model (SLM) and applying quantization, integrating offline STT/TTS, building local RAG, to rigorously validating quality using AI SuperVision and overcoming hardware performance constraints. In this grand finale, Part 10,
Mar 2
![[On-Device AI Chatbot] Part 9: Challenging Performance Limits: Heat, Battery, and Response Speed](https://static.wixstatic.com/media/2ea07e_826bc45db874477090ea018335b34059~mv2.png/v1/fill/w_745,h_559,fp_0.50_0.50,q_35,blur_30,enc_avif,quality_auto/2ea07e_826bc45db874477090ea018335b34059~mv2.png)
![[On-Device AI Chatbot] Part 9: Challenging Performance Limits: Heat, Battery, and Response Speed](https://static.wixstatic.com/media/2ea07e_826bc45db874477090ea018335b34059~mv2.png/v1/fill/w_745,h_559,fp_0.50_0.50,q_95,enc_avif,quality_auto/2ea07e_826bc45db874477090ea018335b34059~mv2.png)
[On-Device AI Chatbot] Part 9: Challenging Performance Limits: Heat, Battery, and Response Speed
Challenging Performance Limits Heat, Battery, and Response Speed In Part 8, we shared how we caught hallucinations and improved response quality using 'AI SuperVision'. While making the model smarter and more accurate is a huge milestone, running it in a real-world smartphone environment (like the Galaxy S25 FE) forces us to confront harsh physical walls: Thermal management, Battery consumption, and Latency limits. Unlike the limitless resources of cloud data centers, a mob
Feb 27
![[On-Device AI Chatbot] Part 8: Catching Hallucinations: Analyzing SuperVision Test Results](https://static.wixstatic.com/media/2ea07e_69fba1e933354148a97a50bbfb2f2dcb~mv2.png/v1/fill/w_745,h_559,fp_0.50_0.50,q_35,blur_30,enc_avif,quality_auto/2ea07e_69fba1e933354148a97a50bbfb2f2dcb~mv2.png)
![[On-Device AI Chatbot] Part 8: Catching Hallucinations: Analyzing SuperVision Test Results](https://static.wixstatic.com/media/2ea07e_69fba1e933354148a97a50bbfb2f2dcb~mv2.png/v1/fill/w_745,h_559,fp_0.50_0.50,q_95,enc_avif,quality_auto/2ea07e_69fba1e933354148a97a50bbfb2f2dcb~mv2.png)
[On-Device AI Chatbot] Part 8: Catching Hallucinations: Analyzing SuperVision Test Results
Catching Hallucinations Analyzing SuperVision Test Results In Part 7, we built an automated testing pipeline that bridged our on-device chatbot app inside a smartphone with the AI SuperVision server on a PC. This enabled an end-to-end flow from prompt injection and answer extraction to automated grading. We finally had an environment capable of running dozens of test cases automatically. So, what kind of report card did our on-device SLM (Gemma-2B based) receive from these
Feb 24
![[On-Device AI Chatbot] Part 7: Building SuperVision: An Automated Chatbot Testing Pipeline](https://static.wixstatic.com/media/2ea07e_22b8a8781b1743cb8aaa018b782ab4da~mv2.png/v1/fill/w_745,h_559,fp_0.50_0.50,q_35,blur_30,enc_avif,quality_auto/2ea07e_22b8a8781b1743cb8aaa018b782ab4da~mv2.png)
![[On-Device AI Chatbot] Part 7: Building SuperVision: An Automated Chatbot Testing Pipeline](https://static.wixstatic.com/media/2ea07e_22b8a8781b1743cb8aaa018b782ab4da~mv2.png/v1/fill/w_745,h_559,fp_0.50_0.50,q_95,enc_avif,quality_auto/2ea07e_22b8a8781b1743cb8aaa018b782ab4da~mv2.png)
[On-Device AI Chatbot] Part 7: Building SuperVision: An Automated Chatbot Testing Pipeline
Building SuperVision An Automated Chatbot Testing Pipeline In Part 6, we explained the background of introducing Testworks' 'AI SuperVision' tool to objectively evaluate the chronic hallucination issues inherent in generative AI. However, to actually apply this tool to our project, we had to overcome a significant technical barrier. Our LLM chatbot operates completely offline "On-device" (inside a smartphone), whereas the AI SuperVision system evaluating it exists in a "PC
Feb 23
![[On-Device AI Chatbot] Part 6: How to Verify AI Quality? (Introduction to SuperVision)](https://static.wixstatic.com/media/2ea07e_38184c3eec5940288ae0fcc2e73f6e2d~mv2.png/v1/fill/w_745,h_559,fp_0.50_0.50,q_35,blur_30,enc_avif,quality_auto/2ea07e_38184c3eec5940288ae0fcc2e73f6e2d~mv2.png)
![[On-Device AI Chatbot] Part 6: How to Verify AI Quality? (Introduction to SuperVision)](https://static.wixstatic.com/media/2ea07e_38184c3eec5940288ae0fcc2e73f6e2d~mv2.png/v1/fill/w_745,h_559,fp_0.50_0.50,q_95,enc_avif,quality_auto/2ea07e_38184c3eec5940288ae0fcc2e73f6e2d~mv2.png)
[On-Device AI Chatbot] Part 6: How to Verify AI Quality? (Introduction to SuperVision)
How to Verify AI Quality? Introduction to SuperVision In Part 5, we explored how to inject our company's proprietary knowledge into the on-device chatbot using Local RAG (Retrieval-Augmented Generation) and Multi-Context Switching. However, equipping the chatbot with knowledge does not immediately solve all problems. "How can we be absolutely sure that this chatbot isn't fabricating answers and is truthfully speaking only about what is in the provided documents?" In Part 6,
Feb 21
![[On-Device AI Chatbot] Part 4: The Ears and Mouth of a Chatbot: On-Device STT/TTS Integration](https://static.wixstatic.com/media/2ea07e_f9b2f825229d4e4b8e86be78ac4fd73b~mv2.png/v1/fill/w_745,h_559,fp_0.50_0.50,q_35,blur_30,enc_avif,quality_auto/2ea07e_f9b2f825229d4e4b8e86be78ac4fd73b~mv2.png)
![[On-Device AI Chatbot] Part 4: The Ears and Mouth of a Chatbot: On-Device STT/TTS Integration](https://static.wixstatic.com/media/2ea07e_f9b2f825229d4e4b8e86be78ac4fd73b~mv2.png/v1/fill/w_745,h_559,fp_0.50_0.50,q_95,enc_avif,quality_auto/2ea07e_f9b2f825229d4e4b8e86be78ac4fd73b~mv2.png)
[On-Device AI Chatbot] Part 4: The Ears and Mouth of a Chatbot: On-Device STT/TTS Integration
The Ears and Mouth of a Chatbot On-Device STT/TTS Integration In Part 3, we explored the optimization process of compressing a massive language model to fit the constrained resources of a smartphone and boosting inference speed using the mobile NPU. Now that we have successfully embedded a fast and smart "brain" inside the device, it is time to give our chatbot the "ears and mouth" it needs to interact naturally with users. In a mobile environment, typing out long texts eve
Feb 20
![[On-Device AI Chatbot] Part 5: A Chatbot That Understands Context: Implementing Local RAG and Multi-Context Switching](https://static.wixstatic.com/media/2ea07e_42172a5ac3454535a81160a2408d0b5b~mv2.png/v1/fill/w_745,h_559,fp_0.50_0.50,q_35,blur_30,enc_avif,quality_auto/2ea07e_42172a5ac3454535a81160a2408d0b5b~mv2.png)
![[On-Device AI Chatbot] Part 5: A Chatbot That Understands Context: Implementing Local RAG and Multi-Context Switching](https://static.wixstatic.com/media/2ea07e_42172a5ac3454535a81160a2408d0b5b~mv2.png/v1/fill/w_745,h_559,fp_0.50_0.50,q_95,enc_avif,quality_auto/2ea07e_42172a5ac3454535a81160a2408d0b5b~mv2.png)
[On-Device AI Chatbot] Part 5: A Chatbot That Understands Context: Implementing Local RAG and Multi-Context Switching
A Chatbot That Understands Context Implementing Local RAG and Multi-Context Switching In Part 4, we gave our chatbot "eyes, ears, and a mouth" by integrating on-device STT and TTS. However, no matter how well a chatbot listens and speaks, it is only half-useful as a business assistant if it doesn't know your specific "domain knowledge"—like internal company regulations or specific product manuals. Because Small Language Models (SLMs) are compact, they do not perform as well
Feb 19
![[On-Device AI Chatbot] Part 3: Core Technologies of Mobile AI: Quantization and NPU Optimization](https://static.wixstatic.com/media/2ea07e_08ed983f9efb45fe9129e06967a91163~mv2.png/v1/fill/w_745,h_559,fp_0.50_0.50,q_35,blur_30,enc_avif,quality_auto/2ea07e_08ed983f9efb45fe9129e06967a91163~mv2.png)
![[On-Device AI Chatbot] Part 3: Core Technologies of Mobile AI: Quantization and NPU Optimization](https://static.wixstatic.com/media/2ea07e_08ed983f9efb45fe9129e06967a91163~mv2.png/v1/fill/w_745,h_559,fp_0.50_0.50,q_95,enc_avif,quality_auto/2ea07e_08ed983f9efb45fe9129e06967a91163~mv2.png)
[On-Device AI Chatbot] Part 3: Core Technologies of Mobile AI: Quantization and NPU Optimization
Core Technologies of Mobile AI Quantization and NPU Optimization In Part 2, we discussed our selection of Gemma-2B as the ideal Small Language Model (SLM) for our project and shared our experiences benchmarking CPU and GPU performance in a constrained smartphone environment. However, the initial tests revealed significant challenges: noticeable latency delays and out-of-memory errors. To run LLMs in real-time on a mobile device held in the palm of your hand—not on a data ce
Feb 18
![[On-Device AI Chatbot] Part 2: Giant Language Models in the Palm of Your Hand: Mobile SLM Selection Strategy](https://static.wixstatic.com/media/2ea07e_7ef19534e8cc4690850ed424d904dee6~mv2.png/v1/fill/w_745,h_559,fp_0.50_0.50,q_35,blur_30,enc_avif,quality_auto/2ea07e_7ef19534e8cc4690850ed424d904dee6~mv2.png)
![[On-Device AI Chatbot] Part 2: Giant Language Models in the Palm of Your Hand: Mobile SLM Selection Strategy](https://static.wixstatic.com/media/2ea07e_7ef19534e8cc4690850ed424d904dee6~mv2.png/v1/fill/w_745,h_559,fp_0.50_0.50,q_95,enc_avif,quality_auto/2ea07e_7ef19534e8cc4690850ed424d904dee6~mv2.png)
[On-Device AI Chatbot] Part 2: Giant Language Models in the Palm of Your Hand: Mobile SLM Selection Strategy
Giant Language Models in the Palm of Your Hand Mobile SLM Selection Strategy In Part 1, we explored how "On-Device AI" is becoming an essential paradigm for solving cloud cost and data security issues. But how can we fit massive Large Language Models (LLMs) with tens or hundreds of billions of parameters—which typically run on massive GPU racks in data centers—into a small smartphone? The answer lies in Small Language Models (SLMs) . In Part 2, we will compare the most nota
Feb 17
![[On-Device AI Chatbot] Part 1: Why "On-Device AI" Now? (Overview)](https://static.wixstatic.com/media/2ea07e_fe141ac84a2c46b8b5daf9987efc1ea7~mv2.png/v1/fill/w_745,h_559,fp_0.50_0.50,q_35,blur_30,enc_avif,quality_auto/2ea07e_fe141ac84a2c46b8b5daf9987efc1ea7~mv2.png)
![[On-Device AI Chatbot] Part 1: Why "On-Device AI" Now? (Overview)](https://static.wixstatic.com/media/2ea07e_fe141ac84a2c46b8b5daf9987efc1ea7~mv2.png/v1/fill/w_745,h_559,fp_0.50_0.50,q_95,enc_avif,quality_auto/2ea07e_fe141ac84a2c46b8b5daf9987efc1ea7~mv2.png)
[On-Device AI Chatbot] Part 1: Why "On-Device AI" Now? (Overview)
Why "On-Device AI" Now? Over the past few years, generative AI, led by services like ChatGPT, has revolutionized our daily lives and workflows. However, behind these powerful AI services lies a common limitation: cloud dependency . The standard architecture—where user queries are sent to cloud servers and the computed results from massive data centers are sent back—inevitably introduces risks such as data privacy breaches, network latency, and exorbitant server maintenance c
Feb 16
![[AX Pro] ⑤ Learning to Walk with AI](https://static.wixstatic.com/media/2ea07e_bc6f4a1789f74859bc1fc9663f9eb266~mv2.jpg/v1/fill/w_1333,h_1000,fp_0.50_0.50,q_30,blur_30,enc_avif,quality_auto/2ea07e_bc6f4a1789f74859bc1fc9663f9eb266~mv2.jpg)
![[AX Pro] ⑤ Learning to Walk with AI](https://static.wixstatic.com/media/2ea07e_bc6f4a1789f74859bc1fc9663f9eb266~mv2.jpg/v1/fill/w_1200,h_900,fp_0.50_0.50,q_90,enc_avif,quality_auto/2ea07e_bc6f4a1789f74859bc1fc9663f9eb266~mv2.jpg)
[AX Pro] ⑤ Learning to Walk with AI
Not Working Alone, But Working 'Together' Through the past four posts, I’ve shared my raw experience with AX Pro, from deployment to real-time operations. As a PO and brand owner, I’ve used countless tools, but few focus as deeply on the essence of 'Human-AI Collaboration' as AX Pro. I didn't just hire a machine that answers questions; I gained a 'Growth-oriented Teammate' that understands our philosophy and evolves under expert guidance. AX Pro main page Three Decisive Mo
Feb 3
![[AX Pro] ④ Operational Evolution: Managing AI with Data, Not Guesswork](https://static.wixstatic.com/media/2ea07e_84e6a80b17474c929a70570cd9ba6b95~mv2.png/v1/fill/w_917,h_688,fp_0.50_0.50,q_35,blur_30,enc_avif,quality_auto/2ea07e_84e6a80b17474c929a70570cd9ba6b95~mv2.png)
![[AX Pro] ④ Operational Evolution: Managing AI with Data, Not Guesswork](https://static.wixstatic.com/media/2ea07e_84e6a80b17474c929a70570cd9ba6b95~mv2.png/v1/fill/w_917,h_688,fp_0.50_0.50,q_95,enc_avif,quality_auto/2ea07e_84e6a80b17474c929a70570cd9ba6b95~mv2.png)
[AX Pro] ④ Operational Evolution: Managing AI with Data, Not Guesswork
Manage AI with metrics, not intuition. AX Pro delivers real-time performance tracking, security with PI Filters, and scalable enterprise AI operations.
Feb 3
![[AX Pro] ③ Taming the AI: "My Chatbot Is Growing Up" (Operations & Feedback)](https://static.wixstatic.com/media/2ea07e_8a73f9acd71242f98ad2f68016dfe489~mv2.png/v1/fill/w_1333,h_1000,fp_0.50_0.50,q_35,blur_30,enc_avif,quality_auto/2ea07e_8a73f9acd71242f98ad2f68016dfe489~mv2.png)
![[AX Pro] ③ Taming the AI: "My Chatbot Is Growing Up" (Operations & Feedback)](https://static.wixstatic.com/media/2ea07e_8a73f9acd71242f98ad2f68016dfe489~mv2.png/v1/fill/w_1200,h_900,fp_0.50_0.50,q_95,enc_avif,quality_auto/2ea07e_8a73f9acd71242f98ad2f68016dfe489~mv2.png)
[AX Pro] ③ Taming the AI: "My Chatbot Is Growing Up" (Operations & Feedback)
Managing the AI's "Puberty" with the Dashboard Deploying the AI isn't the finish line—it's just the beginning. When I first saw the 'Performance Radar' on AX Pro’s main screen, it felt like receiving my son Luke’s report card. Pentagon Metrics : Six key indicators—including Relevance, Accuracy, and Toxicity—are analyzed in real-time and displayed in a radar chart. Overall Score : Our 'Samsung CES 2026' chatbot currently sits at 89 ! A solid grade, but as a PO, I won’t stop
Feb 3
![[AX Pro] ② The 5-Minute Miracle: Building 'Our Own AI' Without a Single Line of Code](https://static.wixstatic.com/media/2ea07e_f459163c86554a718d17ef3bdbd2baa0~mv2.jpg/v1/fill/w_1333,h_1000,fp_0.50_0.50,q_30,blur_30,enc_avif,quality_auto/2ea07e_f459163c86554a718d17ef3bdbd2baa0~mv2.jpg)
![[AX Pro] ② The 5-Minute Miracle: Building 'Our Own AI' Without a Single Line of Code](https://static.wixstatic.com/media/2ea07e_f459163c86554a718d17ef3bdbd2baa0~mv2.jpg/v1/fill/w_1200,h_900,fp_0.50_0.50,q_90,enc_avif,quality_auto/2ea07e_f459163c86554a718d17ef3bdbd2baa0~mv2.jpg)
[AX Pro] ② The 5-Minute Miracle: Building 'Our Own AI' Without a Single Line of Code
Developers? No, I did it all by myself! Usually, adopting an enterprise solution means drafting a request to the dev team, discussing server specs, and spending a month reading complex API docs. However, true to its name—"Expert Augmentation"—AX Pro has lowered the barrier so that experts can handle the tools directly. From the 'Create Account' screen to hitting 'Create New Group' for my first project, the entire process took less time than it takes for a cup of coffee to c
Feb 3
![[AX Pro] ① AI Taking My Job? Please, Take It Already!](https://static.wixstatic.com/media/2ea07e_4c5daba2872e496a9a2dd81c0b7ad362~mv2.png/v1/fill/w_1083,h_812,fp_0.50_0.50,q_35,blur_30,enc_avif,quality_auto/2ea07e_4c5daba2872e496a9a2dd81c0b7ad362~mv2.png)
![[AX Pro] ① AI Taking My Job? Please, Take It Already!](https://static.wixstatic.com/media/2ea07e_4c5daba2872e496a9a2dd81c0b7ad362~mv2.png/v1/fill/w_1083,h_812,fp_0.50_0.50,q_95,enc_avif,quality_auto/2ea07e_4c5daba2872e496a9a2dd81c0b7ad362~mv2.png)
[AX Pro] ① AI Taking My Job? Please, Take It Already!
AX Pro Studio "Is AI really a threat to our careers?" Everywhere you look, it’s all about AI. News headlines warn us that AI will replace human labor, and office hallways are filled with half-joking concerns about when we’ll be packing our bags. But as a PO (Product Owner) and brand owner who never has enough hours in the day, my honest thought was a bit different: "I wish someone—anyone—would just take these repetitive tasks and tedious data sorting off my plate!" Perhaps
Feb 3
AI Supervision 10. The Blueprint for RAG Success: Integrating AI Supervision into Your Architecture
"We built a RAG system, but where exactly does the evaluation tool fit in?" "How do we map the retrieved documents to the actual answer for validation?" The final puzzle piece in LLM service development is Architecture . It’s not just about calling an API; it’s about creating a seamless pipeline that Retrieves, Generates, and Evaluates. In this final article of our series, we present a practical blueprint for integrating AI Supervision into your RAG (Retrieval-Augmented Ge
Jan 20
AI Supervision 9. AI Beyond the Web: Seamless Evaluation with SDKs and Mobile Integration
"Our AI chatbot lives in a mobile app. Do we have to test it on a separate web dashboard?" "Copy-pasting logs from our server to the evaluation tool is tedious." Many AI evaluation tools are stuck in the browser sandbox. However, real users interact with AI in mobile apps, internal messengers like Slack, or complex backend workflows. The gap between the testing environment and the production environment often leads to unexpected bugs. AI Supervision bridges this gap with r
Jan 20
SECURE YOUR BUSINESS TODAY
bottom of page