The latest articles on DEV Community by laya reddy (@laya2211). https://dev.to/laya2211 https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F1739815%2Fa0d274cd-2817-458c-854a-6910bea37600.png https://dev.to/laya2211 en laya reddy Mon, 05 Oct 2026 04:39:18 +0000 https://dev.to/laya2211/build-for-a-friend-an-interactive-dsa-and-algorithm-coach-for-competitive-exam-prep-5e97 https://dev.to/laya2211/build-for-a-friend-an-interactive-dsa-and-algorithm-coach-for-competitive-exam-prep-5e97 <h2> <strong>What I Built</strong> </h2> <p>I built the <strong>GATE CS Visual Dry-Run Tracer</strong> for a close friend who is aggressively preparing for highly competitive core engineering exams like GATE CS, ISRO, and DRDO. </p> <p>They kept losing critical marks on technical algorithm evaluation papers because standard compilers only print out a final code output—they don't show you <em>where</em> your manual dry-run logic diverged at step 3 or 4 of a 10-step execution trace. </p> <p>This app breaks down any complex code block, nested loop, or graph algorithm problem into an interactive step-by-step matrix table. It allows the student to commit their intermediate checkpoint values in a scratchpad before revealing the absolute step-by-step truth, making it a perfect tool for diagnosing logical slips under exam time pressure.</p> <h2> <strong>Demo</strong> </h2> <p>👉 <a href="https://huggingface.co/spaces/laya-03/gate-cs-tracker" rel="noopener noreferrer">Launch the GATE CS Visual Dry-Run Tracer App</a></p> <h2> <strong>Code</strong> </h2> <p>🔗 <a href="https://huggingface.co/spaces/laya-03/gate-cs-tracker/tree/main" rel="noopener noreferrer">View the Application Source Repository</a></p> <h2> <strong>How I Built It</strong> </h2> <p>The application is built completely on top of an open-source architecture:</p> <ul> <li> <strong>The Core Brain:</strong> We utilize the open-weights <strong>GPT-OSS-20b</strong> model, processed through a high-speed open-access Groq API client pipeline.</li> <li> <strong>Interface &amp; UX:</strong> Designed a clean, distraction-free grid-split interface using the open-source <strong>Gradio</strong> framework.</li> <li> <strong>Deployment Infrastructure:</strong> Deployed as a standalone application inside Hugging Face Spaces utilizing the open-access <strong>ZeroGPU</strong> compute tier wrapper.</li> <li> <strong>Prompt Harness:</strong> Engineered custom low-temperature (<code>temperature=0.2</code>) instruction sets to enforce strict, predictable mathematical markdown transformations from natural text input into deterministic tables.</li> </ul> <h2> Why Does Open Innovation Matter? </h2> <p>Open innovation was the only reason this project was possible to ship over a weekend without a corporate budget:</p> <ol> <li> <strong>Infinite Execution Token Capacity:</strong> Tracing complex algorithm loops step-by-step or detailing matrix transitions consumes thousands of context tokens per query. Standard commercial pay-per-token APIs would bankrupt a student using it for daily prep. Open weights on open-access endpoints mean <strong>infinite free dry-runs</strong>.</li> <li> <strong>Deterministic Stability:</strong> Proprietary models frequently drift in behavior overnight due to black-box alignment updates. By grounding our tool in an open-weights standard with zero temperature variations, we enforce static academic formatting required for error-free textbook analysis.</li> </ol> <h2> Prize Categories </h2> <ul> <li><strong>Open Source AI Novelty</strong></li> <li><strong>Build for a Friend</strong></li> </ul> devchallenge weekendchallenge hf26challenge