Loading...
Community presentation
ROLE You are building a 7-slide hackathon pitch deck for the iQOO Hackathon 2026, HealthTech track, Hyderabad City Battle. Team: Git Gang (2 members). This is an IDEA-SCREENING …
Create your own version
Use this visual direction as inspiration for a new Presenton deck.
Use this designCreator prompt
ROLE You are building a 7-slide hackathon pitch deck for the iQOO Hackathon 2026, HealthTech track, Hyderabad City Battle. Team: Git Gang (2 members). This is an IDEA-SCREENING submission — nothing is built yet beyond UI/UX mockups, so do not present unmeasured numbers as fact; frame them as engineering targets. PRODUCT NAME Offbite — an offline-first nutrition and macro logging agent. (Alt names if needed: Kcal, MacroLock) PROBLEM STATEMENT Cloud-based nutrition/voice trackers fail in three specific ways: 1. Network Dependency — cloud NLP fails completely in gym basements, flights, rural trails: exactly the places people most want to log a meal. 2. Mathematical Hallucination — LLM-calculated macros are non-deterministic; they silently corrupt calorie/macro data. 3. Biometric Data Exposure — streaming raw audio and dietary logs to third-party servers creates real privacy/compliance liability. TARGET USER People who need to log meals in places with unreliable connectivity — gym-goers, frequent flyers, trekkers/outdoor athletes. Not "everyone" — anchor every slide to this specific pain point, since it's the actual differentiator vs. HealthifyMe, Cronometer, and MyFitnessPal (all cloud-dependent). SOLUTION ARCHITECTURE — "Two-Brain" system Part A — The Probabilistic Ear (Input Layer) - Llama 3.2 1B-Instruct, INT4 quantized, running on-device via Qualcomm Neural Processing SDK / Android AICore on the Hexagon NPU (Snapdragon 8 Elite Gen 5 — get the Gen 5 right, it's the actual iQOO 15 chip) - Job: NLU only. Converts voice + camera input into a strict JSON schema. Performs zero arithmetic. - Multimodal: camera identifies the dish from a photo of the actual plate (not a packaged-food label — most of the demo food is home-cooked), voice confirms/adjusts portion size in parallel. Part B — The Deterministic Core (Execution Layer) - Local SQLite database seeded from IFCT 2017 (Indian Food Composition Tables, National Institute of Nutrition, ICMR — headquartered in Hyderabad). - IFCT covers 528 raw ingredients but not composite dishes, so add a "recipe layer": common Indian dishes (paneer bhurji, dal, sabzi) mapped to raw- ingredient ratios, built on top of IFCT. - Receives structured JSON from the SLM, does exact lookups and macro math. Math here is deterministic by definition — say this plainly rather than branding it as an "achievement." - Be honest: the real error surface is the SLM's entity extraction, not the math. Don't imply the whole pipeline is error-free — only the arithmetic is. HARDWARE / RUBRIC-ALIGNED DIFFERENTIATORS - NPU offloading: INT4 quantization on the Hexagon NPU instead of CPU Web Workers. Target throughput ~70 tokens/sec — label explicitly as a target, not a benchmarked result. - Thermal gating: heavy tasks (vector indexing, weekly report generation) restricted to active-charging states to respect the vapor-chamber envelope. - Office Kit integration: weekly wellness reports (activity, macro adherence) pushed to the user's PC using Office Kit's real file-transfer and clipboard- sync features. Do not claim a "native Office Kit API" — it isn't publicly documented; describe the actual consumer features instead. - Red Light workflow: because NLU is decoupled from the UI, the team can debug and prompt the local model entirely from the phone during phone-only build time, using Office Kit Screen Mirror to work from a bigger display. JUDGING RUBRIC TO DESIGN AGAINST (confirmed, iQOO Hackathon 2026) End product quality 30% (jury) | Novelty & impact 20% (jury) | Creative phone use — camera/voice/on-device AI 15% (HackTracker device data) | Technical depth 15% (jury) | Office Kit usage 10% (HackTracker device data) | Demo & presentation 10% (jury) DEMO SCRIPT — "The Airplane Mode Test" 1. Toggle iQOO 15 to Airplane Mode on stage (Wi-Fi + Cellular off). 2. Photograph an actual plate of food (e.g. rotis + paneer bhurji) AND speak the portion size. 3. System extracts JSON intent, looks up the recipe-layer + IFCT data, logs exact macros — target under 200ms, stated as a target. 4. Be explicit in the pitch: this proves zero cloud dependency, not accuracy — don't conflate the two claims. DESIGN DIRECTION Dark background (near-black), single orange accent color, white body text, monospace/mono-spaced labels for section tags ("01 / PROBLEM"), clean card- based layout, generous whitespace, consistent header pattern across slides. No gradient bars, no decorative side stripes, no accent lines under titles. SLIDE STRUCTURE (7 slides) 1. Title — Offbite, one-line tagline anchored to the connectivity-loss use case, HealthTech track / Hyderabad City Battle, "Git Gang" credit. 2. Problem — the three cloud-AI failure modes, each tied back to the target user's real situations (gym, flight, trail). 3. Solution — Two-Brain architecture, Probabilistic Ear vs. Deterministic Core, model named explicitly, honest framing of where errors can occur. 4. Runtime pipeline — photo + voice input → NPU extraction → validated JSON → deterministic lookup against IFCT + recipe layer. No dangling/cut-off elements — every step shown must resolve on-slide. 5. Hardware + data — NPU offloading (labeled as target metrics), thermal gating, and the IFCT/NIN-Hyderabad data story as a "Technical Depth" flex. 6. Build strategy — Office Kit's real features (file transfer, clipboard, screen mirror) and the Red Light debugging workflow. 7. Live demo — Airplane Mode Test, step-by-step, with the accuracy vs. no-cloud-dependency distinction stated clearly. OUTPUT FORMAT Generate this as an actual editable slide deck (real text layers, not flattened images) — 16:9, one file, ready to present.
Presentation overview
This community presentation was shared by phimosis metamorphosis as a complete 7-slide example. ROLE You are building a 7-slide hackathon pitch deck for the iQOO Hackathon 2026, HealthTech track, Hyderabad City Battle. Team: Git Gang (2 members). This is an IDEA-SCREENING … The preview lets you review the full sequence in order, rather than judging the design from a single cover image.
Use it as a reference for planning your own deck: notice how the amount of information changes from slide to slide, where visual emphasis appears, and how repeated design choices help the presentation feel connected. Community examples are inspiration, not locked templates, so you can keep the ideas that fit your audience and replace anything that does not.
Learn from the example
Look at the job each slide performs. Identify where the deck introduces its subject, develops the main points, adds evidence or examples, and moves toward a conclusion.
Notice which element attracts attention first and how headings, supporting text, images, and data are separated. Strong hierarchy makes the intended reading order obvious.
Compare spacing, typography, color, and repeated components across the slides. Consistent rules help varied content feel like one presentation instead of unrelated screens.
Make it your own
Begin with your audience and the decision, lesson, or action the presentation should support. Replace the topic, examples, evidence, and visuals with material you can verify, then edit every slide for one clear takeaway. You can borrow the visual direction without copying the creator's wording or message. Presenton will use this community deck as a design reference while you develop content for your own purpose.
