paranoianovel Chapter 31 Joint Review
Chapter 31 Joint Review
The review was scheduled for the forty-second day of the internship, which was also the last day.
The location wasn't the 23rd-floor work area, but rather the main conference hall on the 30th floor of the Blue Bay Communications headquarters building—a floor Zuo Cheng had never been to before. As soon as the elevator doors opened, display boards showcasing Blue Bay Communications' technological milestones over the years hung on both sides of the corridor, the furthest dating back to the company's founding twenty years ago.
The conference room was large, with a long, rectangular judging table spanning the entire room, where nine people sat.
Zuo Cheng glanced at the nameplates on the judges' panel.
In the center is Zhou Henian. To his left sits the chief engineer of the Sky Dome project team—the one Zuo Cheng had spotted with the technological radar. To his right are two people Zuo Cheng doesn't recognize, their nameplates reading "Director of the Communication Payload Research Laboratory, Blue Star Aerospace Academy" and "Professor of the Satellite Communication Laboratory, National University of Defense Technology, China."
External experts.
The remaining five were members of the Blue Bay Communications Technology Committee, all of them senior engineers.
The nine-person judging panel was even more impressive than Zuo Cheng had anticipated. This wasn't an intern's assignment—it was an assessment of the technical feasibility of the first phase of the Sky Dome project.
Four interns sat across from the review table, each with a printed proposal document and a laptop in front of them.
According to the order of the draw, Cheng Yuan was first.
His presentation was as solid as ever—the PowerPoint slides were neatly arranged, his speaking pace was well-controlled, and the architecture diagram of the deep learning model was beautifully drawn. He had indeed put in a lot of hard work, training a fairly impressive neural network model in six weeks, and achieving a good level of prediction accuracy in standard scenarios.
But the judges won't just look at the standard scenario.
The director of the Aerospace Science and Technology Academy asked the first question: "How does your model perform in scenarios outside the training data distribution? For example, extreme weather conditions not included in the training set."
Zuo Cheng mentally repeated Cheng Yuan's answer for him: "Improve the model's generalization ability through data augmentation and transfer learning techniques."
Sure enough, Cheng Yuan said it almost word for word.
But the director pressed further, "What about the quantitative evaluation of generalization ability? Have you conducted blind tests on entirely new scenarios outside the training set?"
Cheng Yuan paused for two seconds. "We haven't done a full blind test yet; there's not enough time. But theoretically—"
"Theoretical generalization and practical generalization are two different things," the director interrupted him, his tone not harsh but clear. "The biggest risk of deep learning models is getting out of control on unseen data, and satellite communication frequently encounters extreme situations outside the training set. You must address this issue later."
Cheng Yuan nodded and calmly returned to his seat. But Zuo Cheng noticed that he tightened his grip on the pen slightly.
The second one is Tang Xu.
Tang Xu's antenna array solution was solid, his beamforming algorithm design was clear, and the simulation data presented on-site was very convincing. He answered several technical questions from the judges well, especially regarding the optimization of beam switching delay—he reduced the switching delay from 15 milliseconds in the traditional solution to 3 milliseconds, a figure that made the chief engineer raise an eyebrow noticeably.
The third is Lin Ke.
Her spatial channel modeling proposal was the most academic of the four—it included extensive physical derivations and mathematical proofs. Several judges read through her documents very slowly because of the high information density. During the defense, she debated the details of ionospheric parameterization modeling with a professor from the National University of Defense Technology for five minutes. Finally, the professor said, "Your approach is more refined than our group's current solution," and Lin Ke calmed down.
The fourth is Zuocheng.
He stood up and walked to the projection screen, but did not open the PowerPoint presentation.
"Good morning, esteemed judges. I am Zuo Cheng, and I am responsible for the satellite-to-ground link channel estimation and prediction algorithm. Before we begin, I would like to show you some data."
He opened his laptop and brought up a real-time simulation demonstration program—built with Fang Ze's remote assistance—that could display the working status of the two-layer prediction architecture in real time under any set satellite orbit, weather conditions, and ground environment.
An animation of a virtual satellite passing over the surface of Earth appeared on the screen. The ground area below the satellite was marked with different colors—green for sunny days, blue for cloudy days, and red for heavy rain. As the satellite moved, the channel state curve scrolled in real time on the right side of the screen, with the predicted and actual values almost completely overlapping.
Then Zuocheng switched the weather conditions to extreme rainstorm.
The red area expanded, and the channel curve fluctuated violently—but the prediction line only wobbled for less than a second before re-aligning with the true value.
He then cut off the solar storm. The ionospheric disturbance caused a significant jump in the channel, and the prediction line was corrected in a very short time.
The meeting room was silent for a few seconds.
The chief engineer unconsciously touched his chin.
"This is a real-time demonstration of the two-layer prediction architecture." Zuo Cheng turned off the animation and turned to the review panel. "The lower layer makes deterministic predictions based on orbital mechanics and meteorological data, while the upper layer uses an adaptive tracking algorithm for stochastic compensation. The coordination between the two layers is ensured by a millisecond-level synchronization mechanism—this synchronization mechanism can also output a high-refresh-rate channel prediction interface, supporting beamforming decisions for antenna arrays."
He paused for a moment, his gaze sweeping across the judges' panel.
"Next, I will present the detailed contents of the plan."
Then, in twelve minutes, he explained the theoretical framework, mathematical derivation, simulation results of seven extreme scenarios, and engineering implementation plan of the two-layer architecture. Not a single word was wasted. Each technical point was accompanied by screenshots of simulation data, and the way the data was presented—which Zuo Cheng and Shen Yue had communicated remotely, drawing on the visual logic of the New Talent Cup roadshow—was so clear that the judges could understand it without needing to ask further questions.
Q&A session with the judges.
The first question came from the chief engineer: "Your underlying forecasting relies on real-time meteorological data input. What happens if the meteorological data source is interrupted?"
"The underlying layer has a built-in degradation mode," Zuo Cheng replied. "When meteorological data is interrupted, the underlying layer automatically switches to pure orbit prediction—the accuracy will decrease, but the adaptive compensation capability of the upper layer is sufficient to cover the fallback, ensuring that the system will not crash due to the failure of a single data source. The detailed design and simulation verification of the degradation mode are available on page 47 of the solution document."
The chief engineer turned to page forty-seven, looked at it for half a minute, and nodded.
The second question came from a professor at the National University of Defense Technology: "Your meteorological interpolation module uses antenna beam scanning data as auxiliary constraints. Was this design completed independently or in collaboration with Tang Xu?"
Zuo Cheng did not hesitate.
"This was a collaborative effort. The insight that beam scanning data can reflect the spatial distribution of rainfall was proposed by Tang Xu, the mathematical framework of the interpolation algorithm was designed by me, and the data preprocessing was done by Tang Xu. I have noted this in the acknowledgments section of the project document."
The professor flipped to the acknowledgments page to confirm, then glanced at Tang Xu and nodded slightly.
The third question comes from Zhou Henian.
He remained silent throughout the entire review session, quietly flipping through documents and taking notes. Only after all the technical questions had been asked did he raise his head.
"Zuocheng, one last question."
"Please speak."
"Besides satellite-to-ground links, in which other scenarios can your two-layer prediction architecture be used?"
This question isn't testing technical skills, it's testing perspective.
Zuo Cheng thought for three seconds.
"This framework can be applied to any prediction problem with a structure of 'deterministic skeleton plus random noise'." His voice wasn't loud, but every word was clear. "Inter-satellite links—the relative motion between satellites is deterministic, and the random perturbations in the inter-satellite channel are caused by space particle radiation. Deep space exploration—the orbital relationship between Earth and the probe is deterministic, and the random perturbations in the deep space channel are caused by the solar wind and the interplanetary medium. Even beyond the field of communications—trend forecasting in financial markets, short-term forecasting in weather systems, and trajectory planning for autonomous driving—the underlying logic can all be reduced to 'layered processing of deterministic and random factors.'"
The meeting room fell silent for a few seconds.
A very slight smile appeared at the corner of Zhou Henian's mouth.
"Okay. The review is over. The four students can go back and rest now. We will notify you of the results today."
The four of them walked out of the conference hall. In the corridor, Tang Xu let out a long sigh, Lin Ke stretched her stiff shoulders, and Cheng Yuan walked at the very back with a blank expression.
Zuo Cheng stood at the elevator entrance when his phone vibrated.
A message from Yu Ying.
"Today is the judging day, right? How was it?"
Zuo Cheng typed a few words, then deleted them. After thinking for a moment, he replied, "Wait for the result."
Yu Ying replied instantly: "Then I'll wait for your good news."
Then she sent another message: "By the way, the official published version of the paper has been sent to the college. Professor Lin kept two copies for us. Remember to pick them up when you get back."
Zuo Cheng looked at the last sentence, "Come back," and the corners of his mouth curved slightly.
come back.
Yes, it will be soon.
At 5 p.m., Han Zhe sent a message.
"Joint review results: All four students' proposals have passed. Detailed scores and reviewer comments have been sent to their respective email addresses. Please check your inboxes."
Zuo Cheng opened his email.
Overall score: 97.
Nine judges, the lowest score was 92 and the highest score was full marks.
The person who gave it a perfect score was Zhou Henian.
A burst of golden light exploded in my consciousness.
[Main Quest Chain - Breaking the Communication Impasse - Stage Four: Completed!]
[Joint Review Evaluation: Excellent (97 points, highest rating)]
[Reward for Stage Four: Unlock the "Satellite Communication Basics" blade ✓, +15 points]
[Main Quest Chain - Communication Breakthrough: The Entire Chain Completed!]
[Blockchain completion rewards are being distributed]
[Unlock Branches: Communications Engineering ✓✓✓]
[Bonus Points +20]
[Technology Tree Evolution - Phase One Unlocked]
Golden light surged up from the roots at the base of the technology tree, climbing along the main trunk and illuminating every leaf and branch. The branch labeled "Communications Engineering" changed from dark gold to brilliant gold in the light, and then lit up steadily.
The entire technology tree trembled in Zuo Cheng's consciousness, as if something was awakening.
[Technology tree evolving...]
[Emerging stage → Growth stage]
The outline of the tree canopy emerged in the darkness for the first time—blurred, enormous, and far beyond his current vision. Countless gray branches hung down from the canopy, each marking a territory he had yet to explore.
He vaguely saw a few names—Internet of Things, Artificial Intelligence, New Energy…
Then the light faded, and everything returned to calm.
The tech tree has evolved to: Growth Stage
[Upgraded Technology Radar: Range 30 meters, Cooldown Time 48 hours]
[New Capabilities Unlocked: Technological Integration]
[Task Reward Multiplier: ×1.2]
Current points: 88
[Current blade count: 9]
Current branch: 1
Zuo Cheng sat on the edge of his dormitory bed, staring at the newly transformed technology tree in his mind.
Eighty-eight points. Nine leaves. One branch.
There is also a brand new capability – technological integration.
Compared to the poor student with only 237 yuan in his card six months ago, he is now in a completely different position.
But this is only the beginning.
The somber outlines of the branches in the tree crown silently reminded him that communications engineering was just the first branch on this tree. He still had a long way to go before the whole tree could grow into a towering giant.
Zuo Cheng turned off the control panel, picked up his phone, and sent a message to Yu Ying.
"Passed. Ninety-seven points."
Ten seconds later, Yu Ying replied.
Only two words: "So good."
The same two words as on the day of the final.
Zuo Cheng smiled.
He got up and walked to the window, pushed it open, and the night wind blew in, carrying the chill of early spring and the lingering smell of firecrackers drifting from somewhere in the distance.
It's time to go back to school.