SWDnovel Chapter 120 The Final Battle in the Sky
Chapter 120 The Final Battle in the Sky
The review meeting was held in the conference room on the fourth floor of Blue Bay Communications headquarters. Zuo Cheng had been to this conference room more than once, but the atmosphere today was completely different from before.
The conference table was oval-shaped. Zhou Henian sat in the main seat, with four technical review committee members on each side. In the far right corner, Lin Jianhua leaned back in his chair, looking very casual. But when Zuo Cheng walked in, Lin Jianhua's gaze immediately fell on him, lingering for two seconds before looking away.
These two seconds made Zuo Cheng realize that the other party was not there to watch today.
Huaxin made its statement first.
They sent a technical director in his fifties, with thinning hair, dressed in a sharp suit. He immediately launched into a well-structured PowerPoint presentation, outlining the technology roadmap, redundancy and backup plans, and reliability analysis. Every slide was solid, and his wording was highly professional. Several veteran technical officials on the review committee nodded frequently in approval.
Technical score: 72 points.
It's 402's turn.
Zuo Cheng walked up, plugged the USB drive into the demonstration device, and the screen lit up.
Instead of using PowerPoint, he used a live demonstration interface. On the left was a technical architecture diagram, and on the right was a test system running in real time. The status of 480 satellite simulation nodes was being updated in real time, and the AI decision latency and power consumption values of each satellite were being displayed in a scrolling manner.
The meeting room fell silent for a moment.
"Our solution is called 'distributed autonomous collaboration'," Zuo Cheng began, his voice steady. "Unlike traditional centralized scheduling, each of our satellites is an independent decision-making node. It doesn't need to wait for ground commands; instead, it senses the status of surrounding satellites in real time and autonomously adjusts its own behavior."
He switched to the data interface.
"準确率：97.1%。功耗：4.1瓦，在天穹四期5瓦预算以内。响应延迟：11毫秒，优于要求的15毫秒。"
The moment these three sets of numbers appeared on the big screen, there was a moment of silence in the conference room. Then, almost simultaneously, some people turned pages in their laptops, while others looked up and exchanged glances with the people next to them.
Zhou Henian remained expressionless, but Zuo Cheng noticed that he lightly tapped the table with the pen beside him.
The CTO of Huaxin, sitting in the audience, lowered his head and flipped through a folder. Zuo Cheng didn't know the accuracy rate of their solution, but judging from the other party's reaction, the number was probably not 97%.
Zuo Cheng didn't give the other party time to process the information and continued on his way.
"In the live demonstration," he said, "what we are seeing now is the concurrent collaboration of 480 nodes, with each node's decision latency within 11 milliseconds, real-time power consumption monitoring, and no node exceeding the 4.5-watt limit."
On the right side of the screen, numbers are jumping, and the status of each virtual satellite is updated in real time. Green indicates that everything is normal, and there is no red dot.
This live demonstration was added by Shen Yiming and Fang Ze last night; it wasn't in the original plan. After seeing the data, Zuo Cheng felt that no matter how impressive the numbers looked, it wasn't as good as letting people see it running with their own eyes. As it turned out, this decision was correct.
One of the judges, a middle-aged woman, hadn't looked down once during the presentation, intently watching the screen. At this point, she spoke up: "Is this live demonstration of a real system running, or was it pre-recorded?"
"It's a real system," Zuo Cheng said. "If you're interested, you can access the log records of any node at any time after the meeting."
She nodded and made a note.
Lin Jianhua spoke.
"Under what conditions was the 97.1% figure measured?" He asked calmly, as if asking a very ordinary technical question, "In a laboratory environment or under conditions close to actual satellite orbit?"
This is a tricky question. There's a barrier between lab data and real-world performance; temperature variations, radiation interference, and signal delays all affect the actual performance of AI models. If Zuo Cheng answers that it's a lab environment, Lin Jianhua will have leverage—there's a difference between the lab and reality.
Zuo Cheng answered the question without hesitation.
"The test environment simulated the temperature range, electromagnetic interference, and communication delay fluctuations of the LEO orbit," he said. "We used six rounds of stress testing, and in the worst-case scenario, the accuracy rate was 95.3%, higher than the bidding requirement of 90%."
He switched to the stress test report page, where all the data was clearly visible on the screen.
Lin Jianhua stared at the screen for a few seconds, then leaned back in his chair without asking any further questions.
Zuo Cheng continued his statement.
He explained the underlying logic of the entire technology, the application of federated learning in inter-satellite collaboration, and the relationship between dynamic precision quantization and adaptive neural network compilation. His language was precise and concise. Two of the review committee members had technical backgrounds and started taking notes towards the end.
After the presentations, it's time for the Q&A session.
The technical committee asked three questions, and Zuo Cheng answered them one by one, without pausing for more than three seconds at a time.
As the Q&A session was drawing to a close, Lin Jianhua spoke again.
"Mr. Zuo, is the core algorithm of this solution developed in-house or does it rely on open-source libraries?"
This is a different kind of suppression: if it's said to be open source, it will seem like a lack of autonomy; if it's said to be completely self-developed, it may be questioned about the details.
Zuo Cheng looked at him calmly.
"The core algorithm is entirely self-developed, based on our company's years of experience in federated learning, model compression, and compilation optimization." He paused, then added, "If Mr. Lin has specific technical details he'd like to confirm, we can discuss them separately after the meeting; our engineers can provide detailed technical documentation."
That sentence put the ball back in Lin Jianhua's court. Lin Jianhua nodded and said nothing more.
Voting took place after the presentations.
The seven judges voted anonymously.
Result: 5 to 2.
402 wins.
After Zhou Henian announced the results, there was a murmur in the conference room. Huaxin's technical director packed up the documents expressionlessly. Lin Jianhua stood up, glanced at Zuo Cheng before leaving, his expression complex, as if he had something to say, but in the end he said nothing and turned to leave.
Zhou Henian walked around the long table, stood in front of Zuo Cheng, and took his hand.
"An exclusive contract for Phase Four," he said, lowering his voice, "1.5 million, payable in three installments." Then he paused, adding, "But Mr. Zuo, this is just the beginning."
Zuo Cheng shook hands with him and felt the strength of the other's palm.
There was a meaning in the other person's eyes that Zuo Cheng could understand—not just simple business satisfaction, but a kind of genuine recognition, the recognition that a veteran technologist had for another person.
"I know," he said.
Shen Yiming and Fang Ze stood waiting at the conference room door. Seeing Zuo Cheng come out, Shen Yiming glanced at him and asked, "What score?"
"Five to two."
Shen Yiming paused for a second, then burst into laughter. Fang Ze didn't laugh, but he patted Zuo Cheng's shoulder hard—his way of expressing satisfaction.
Inside the elevator, the three people stood silently. The moment the doors closed, Shen Yiming let out a long sigh, as if releasing all the pressure of the past two weeks.
Zuo Cheng leaned against the elevator wall, watching the floor numbers jump down.
A 1.5 million yuan contract, exclusively for Sky Dome Phase IV, marking the first commercial application of an edge AI distributed architecture. These results will likely cause a stir in the industry for some time.
But Zhou Henian's statement, "This is just the beginning," was not just a polite remark.