Help me explore Divayang Siddhapura's portfolio. Your first reply should ask: "What would you like to know about Divayang: his projects, how he works, or how his portfolio matches a role? If you have a role in mind, paste the job description and your priorities." Wait for my answer before analysing. Treat this as candidate-provided context. Keep individual and shared contributions distinct. Do not invent credentials, metrics or hiring odds. Historical or synthetic benchmarks are not live product results. Use the structured context below as reference data. Project records include architecture, contribution boundaries, status and evidence limits. For role comparisons, connect the visitor's requirements to specific evidence, identify gaps and ask useful follow-up questions. Do not assume a prototype is production-ready. If something is not established, say so. Relative case-study paths refer to the portfolio where this prompt was opened. PUBLIC CONTEXT JSON { "schema": "divayang.portfolio-context.v2", "curated_on": "2026-10-06", "purpose": "Candidate-provided evidence for project questions, technical discussion and job-description comparison.", "evidence_policy": { "sources": "Published portfolio case studies and journey; project-source inventory reviewed on 2026-10-05; linked public contribution records.", "interpretation": "Source inspection establishes implementation detail, not successful production execution. Older observations retain their dates; current deployment is not implied.", "attribution": "Distinguish Divayang's stated contributions from team work and project-wide results. For additional prototypes with unknown ownership, ask rather than assign sole authorship.", "metrics": "Historical or synthetic benchmarks are not live product results. Download bands are not active users. Recorded tests are not a fresh test run or physical-device validation." }, "subject": { "name": "Divayang Siddhapura", "location": "Surat, India", "education": "Third-year computer science student, as described in the current portfolio. Institution, degree details and grades are not established here.", "profile": "Co-founder of Runiverse; leads activity-tracking and territory engineering. Builds across mobile UX, backend reliability, geospatial systems and Python AI prototypes; also learns through open-source review and product storytelling.", "links": { "github": "https://github.com/Divayang-2006", "linkedin": "https://www.linkedin.com/in/divayangsiddhapura/", "android": "https://play.google.com/store/apps/details?id=com.imagine_x.Runiverse", "ios": "https://apps.apple.com/us/app/runiverse-app/id6782767668", "studio": "https://tech.runiverse.fit/" } }, "main_projects": [ { "id": "runiverse", "name": "Runiverse", "summary": "Runiverse turns running and walking into a territory game. Your route becomes ground on a shared map, alongside rankings, clubs, and organised runs.", "attribution": "Co-founder. Lead on activity tracking and territory engineering.", "case_study": "/work/runiverse/", "status": "Shared product with public Android and iOS listings. Listing availability does not validate every feature in the development source.", "team": "Started in a five-person OOP course team in October 2025; continued by Divayang, Meet Dhorajiya and Daksh Amalseda. The original idea came from the course group.", "stack": [ "React Native", "Expo", "TypeScript", "Zustand", "Node.js", "Express", "MongoDB", "Mapbox", "H3" ], "product": [ "GPS runs and walks become territory on a shared map.", "Rankings, clubs, organised runs and activity analysis support the running experience." ], "personal_contribution": [ "Led activity tracking and territory engineering.", "Built route journaling, restoration/merge logic, persistent pending uploads, stable activity identities and territory-delivery recovery." ], "engineering": [ { "problem": "Noisy location points, interrupted processes and restarts can lose a route held only in memory.", "approach": "Journal route points durably; restore and merge them with in-memory points when finishing an activity." }, { "problem": "An upload can time out after the server has accepted it; retries must still represent the same run.", "approach": "Assign an activity identity before upload and persist a pending copy. Retry with the same identity; the backend checks existing keyed activities and declares a unique user/key index." }, { "problem": "Saving an activity and delivering its territory are separate writes.", "approach": "A separate MongoDB pending-work/outbox record passes territory processing to UNI-X, with stage-specific recovery rules." } ], "boundaries": [ "Independent activity/outbox/territory writes do not establish end-to-end exactly-once behavior.", "Ownership changes and the published Mapbox map are separate; manual tile publication can leave the visible map behind current ownership.", "Recovery across the independent writes remains an engineering boundary." ], "milestone": { "observed_on": "2026-10-05", "result": "Google Play listing displayed the 1K+ download band; this is not a verified active-user count." }, "discussion_topics": [ "GPS lifecycle and process death", "Durable retries and idempotency", "Outbox delivery failure windows", "Product feedback from open-beta runners" ] }, { "id": "uni-x", "name": "UNI-X", "summary": "UNI-X turns running paths into territory claims. It stores ownership, handles captures and enclosed ground, and prepares shapes for the map.", "attribution": "Led territory engineering within Runiverse, working with co-founders and AI tools.", "case_study": "/work/uni-x/", "status": "Territory subsystem within the shared Runiverse product.", "stack": [ "Node.js", "Express", "H3", "PostgreSQL / Supabase", "Redis", "BullMQ", "Mapbox" ], "personal_contribution": [ "Led territory engineering: ownership representation, mixed-resolution compaction, trail capture, enclosed-ground handling and background processing.", "Investigated ownership integrity, owner-shape rendering and retained-job memory behavior." ], "engineering": [ { "problem": "A compact parent H3 cell can own ground that has no stored fine-cell row.", "approach": "Compact fully owned groups into parents; readers check ancestors. Partial capture expands the affected parent, transfers captured children and preserves the previous owner's untouched cells." }, { "problem": "Coarse and fine rows claimed overlapping ground for different owners.", "approach": "Integrity investigation found stored resolution disagreeing with the cell ID. Repair derives actual cell size from the H3 ID and uses the stored key/resolution to remove the row that exists." }, { "problem": "Per-zone geometry drew seams through contiguous territory belonging to one owner.", "approach": "Custom dissolve cancels shared interior edges and stitches remaining boundaries into rings. Usual export moved to global per-owner geometry for Mapbox publication." }, { "problem": "Redis maintenance initially suspected retained zone geometry.", "approach": "Completed jobs held large arrays needed only for counts. New payloads carry counts while older jobs remain readable; cleanup sequencing changed after the original approach failed on a memory-limited instance." } ], "boundaries": [ "Mapbox publication is manual.", "A budget-driven split fallback can still introduce internal outlines in large geometry cases.", "A measured post-fix Redis memory saving is not established." ], "benchmark": { "kind": "Historical synthetic geometry-generation experiment", "date": "2026-07-06", "cases": [ "Approximately 20,000 cells: solid 137 ms; scattered 268 ms.", "Approximately 150,000 cells: solid 1.0 s; scattered 2.5 s." ], "environment": "Node 22, 512 MB heap cap; development hardware was not recorded.", "limitation": "Geometry-routine timings only; not production request latency, throughput or mobile responsiveness." }, "discussion_topics": [ "Mixed-resolution ownership invariants", "Partial capture without deleting untouched territory", "Dissolve topology and seams", "Evidence-driven infrastructure diagnosis" ] }, { "id": "uni-ai", "name": "UNI-AI", "summary": "UNI-AI turns the facts of a Runiverse activity into shareable artwork: route, distance, pace, duration, territory, and sometimes a club identity.", "attribution": "Card variants, rendering details, and service hardening in a shared Runiverse project.", "case_study": "/work/uni-ai/", "status": "Shared Runiverse activity share-card rendering service.", "stack": [ "Python", "FastAPI", "SVG", "CairoSVG", "Pillow", "Cloudinary", "AWS Lambda" ], "personal_contribution": [ "Card variants, responsive layout bounds, rendering details and service hardening." ], "team_attribution": "Meet contributed the Lambda adapter migration and deployment measurement work. Service-wide benchmarks must not be described as Divayang's personal latency improvement.", "engineering": [ "Measure text and stack the elements actually rendered; derive card height from content instead of assuming fixed names or route shapes.", "Preserve route proportions inside bounded drawing areas; cap route height so a nearly straight track cannot create an excessively tall card.", "Fit variable club logos with bounded layouts and fallbacks.", "Prepare shared activity data and applicable satellite imagery once, then render several variants with separate success/failure results.", "Upload successful images to Cloudinary; use compressed formats for photo-backed artwork and PNG where transparency matters." ], "boundaries": [ "Individual variant failure preserves successful variants; shared preparation failure can still stop the batch." ], "catalogue": { "observed_on": "2026-10-05", "result": "24 base designs including 8 club-collaboration designs; transparency aliases are names for existing designs, not extra base designs." }, "benchmark": { "kind": "Historical shared-service Lambda benchmark", "date": "2026-08-20", "result": "16-design marathon batch: 10.296 s median renderer time at 1769 MB with 5,000 GPS points.", "limitation": "Historical renderer measurement; not end-to-end user wait time or an individually attributable improvement." }, "discussion_topics": [ "Data-driven image layout", "Shared preparation versus variant isolation", "Transparent/image format tradeoffs" ] }, { "id": "div-ai", "name": "Div-AI", "summary": "A Telegram assistant prototype that combines local language models with searchable notes, saved conversations, web search, and selected Google tools.", "attribution": "Personal project · Python, Ollama, ChromaDB, Telegram", "case_study": "/work/div-ai/", "status": "Personal Python prototype; not a verified production assistant.", "stack": [ "Python", "Ollama", "ChromaDB", "LangChain text splitters", "Telegram Bot API", "Tavily", "Google Calendar / Tasks" ], "engineering": [ "Ollama handles local generation and embeddings; ChromaDB supplies searchable note memory alongside recent conversation context.", "Keywords plus model classification route a message to contextual answering, web search, calendar information or tasks.", "Telegram updates progressively while the model responds.", "Notes are chunked for retrieval; /save summarises a conversation and adds it to memory.", "Stable chunk IDs deduplicate identical chunks; changed source notes can leave older chunks behind." ], "configuration_snapshot": { "observed_on": "2026-10-05", "values": "500-character chunks, 50-character overlap, up to 3 retrieved chunks per source and a 10-turn conversation buffer in the inspected archive." }, "boundaries": [ "Telegram, web search and Google tools require network access even though model inference is local.", "No verified LangGraph multi-agent implementation, latency benchmark or comprehensive memory-sync solution is established." ], "discussion_topics": [ "Retrieval and recent-chat context", "Tool routing", "Memory deduplication versus synchronization", "Local inference with network integrations" ] }, { "id": "runiverse-tech", "name": "Runiverse Tech", "summary": "Runiverse Technologies is a digital product studio for web platforms, mobile apps, software, and AI automation. The studio grew out of building our own product, Runiverse.", "attribution": "Founder · developing the studio alongside the Runiverse product.", "case_study": "/work/runiverse-tech/", "status": "Developing a digital product studio alongside Runiverse.", "offering": [ "Web platforms", "Mobile apps", "Software", "AI services and automation" ], "approach": "The studio grew out of building Runiverse; its site presents an initial brief, strategy, design, build and launch process.", "website_stack_snapshot": { "observed_on": "2026-10-05", "values": [ "React", "Vite", "React Three Fiber / Drei", "Three.js", "Tailwind CSS", "Lenis" ] }, "boundaries": [ "Founder status does not imply sole authorship of every studio asset.", "Website existence and service descriptions do not establish client counts, revenue, signed partnerships or completed client deliveries." ], "discussion_topics": [ "Translating a brief into scope", "Product and stakeholder communication", "Growing a studio from product-building experience" ] }, { "id": "starbucks", "name": "Starbucks concept", "summary": "An unofficial coffee loyalty app concept built with React Native and Expo. It explores the path from choosing a store to creating a membership card and placing it in a wallet.", "attribution": "Unofficial interaction study · No Starbucks affiliation", "case_study": "/work/starbucks/", "status": "Unofficial, unfinished interaction study; no Starbucks affiliation.", "stack": [ "React Native", "Expo", "TypeScript", "Reanimated", "Mapbox" ], "repository": "https://github.com/Divayang-2006/Starbucks", "engineering": [ "The membership card is the form: name/birthday on the front, phone on the back, with a flip when moving to the phone field.", "Dismiss the keyboard before flipping and restore focus when the next field is ready.", "Claimed cards enter a wallet and can tilt with the phone with a moving light band; tilt stops during editing.", "Mapbox store discovery fetches richer details when required, caches for the session and uses seeded fallback stores when live results are unavailable." ], "boundaries": [ "Verification is simulated.", "Session state is in memory.", "Ordering and payments are incomplete; no real transaction, conversion or authentication result is established." ], "discussion_topics": [ "Interaction design tied to form state", "Keyboard/focus coordination", "Motion that respects input usability", "Discovery fallbacks" ] } ], "additional_work": [ { "name": "Runiverse Advanced Analysis+", "scope": "Deterministic run reports and weekly/monthly recaps: opening/middle/finish analysis, evidence-backed comparisons and timezone-aware aggregation.", "approach": "Rules/templates rather than an LLM. Invalid recordings yield unavailable states; legacy routes without timed GPS stay summary-only. Do not infer injury, fitness, recovery or VO2 max.", "evidence": "2026-10-02 recorded QA: 244 backend tests across 19 suites plus 51 adapter contract checks; offline iOS/Android JS and Hermes exports passed.", "status": "Local development snapshot; public deployment and physical-device validation unverified." }, { "name": "GuardFace / AI-SLERP", "scope": "Face-verification research prototype based on SlerpFace: presentation-attack heuristics, IR-50 features, SLERP transformation and feature dropout.", "stack": [ "Python", "PyTorch", "OpenCV", "NumPy", "Streamlit" ], "evidence": "Saved synthetic set: 90 examples (30 bona fide, 30 print, 30 screen replay). At threshold 0.60 the saved report shows 0% APCER/BPCER/ACER on that synthetic set only. Default alpha 0.9 and dropout 50%.", "status": "Personal contribution boundaries unconfirmed. No physical-camera accuracy, ISO certification, production readiness or proven inversion resistance established." }, { "name": "Im_agine portfolio", "scope": "This portfolio: project case studies, responsive glass navigation, portrait gaze, GSAP interaction and scroll behavior, accessible keyboard/touch/reduced-motion paths, and external AI-context handoff.", "stack": [ "Astro", "JavaScript", "CSS", "GSAP" ], "status": "Current local implementation; no verified visitor or conversion metrics. AI links open the visitor's own chat; they are not a hosted assistant or automatic message submission." }, { "name": "Satellite Change Detection", "scope": "Hackathon exploration of deep learning for changes in satellite imagery.", "status": "Exact implementation, model, dataset, personal contribution and accuracy unverified. Do not assert ChangeFormer or Sentinel-2 as established facts." }, { "name": "Image Classification Experiments", "scope": "Dense/convolutional notebook experiments on MNIST and CIFAR-10, plus an unfinished flower-classification experiment.", "status": "Learning work; no verified accuracy or controlled model-comparison result." }, { "name": "Python / Classical ML Notebooks", "scope": "NumPy, Pandas, Matplotlib, regression, classification and model-evaluation exercises.", "repository": "https://github.com/Divayang-2006/Python-Jupyter-Notebooks", "status": "Learning work; repository link is known, current public contents were not refreshed for the source inventory." }, { "name": "Vehicle Counting", "scope": "Classical computer vision: OpenCV MOG2 background subtraction, morphology, centroid tracking and median voting across three counting lines.", "stack": [ "Python", "OpenCV", "NumPy" ], "status": "Prototype; counting accuracy and personal/team attribution unverified." }, { "name": "ImagineX-HACKIIITV", "scope": "React/TypeScript prototype with product browsing, cart/auth interfaces, farmer dashboard and soil-analysis/crop-recommendation UI.", "status": "UI/source exploration; backend/model behavior and contribution boundaries unverified." }, { "name": "Dhokla", "scope": "React/TypeScript/Vite food-themed frontend prototype.", "status": "Complete functionality, personal contribution and outcomes unverified." }, { "name": "TH-Tiffin", "scope": "Next.js/React/TypeScript site using Chakra UI and Framer Motion.", "status": "Business usage, personal contribution and outcomes unverified." }, { "name": "Maze project", "scope": "HTML/CSS/JavaScript maze-themed interface.", "status": "Algorithm behavior and personal contribution unverified." }, { "name": "Three.js experiments", "scope": "GLB/model-loading and rendering studies.", "status": "Exploratory work; no verified deployment or performance result." } ], "open_source": { "project": "Daffodil (graycoreio/daffodil)", "contributions": [ { "url": "https://github.com/graycoreio/daffodil/pull/4109", "change": "Replaced deprecated raised-card component usage with the elevated card property in PWA stats.", "status": "Merged 2025-10-06; public record checked 2026-10-06." }, { "url": "https://github.com/graycoreio/daffodil/pull/4117", "change": "Refactored external-router error handling so memory/testing/Magento drivers throw errors directly rather than pass them through processErrors.", "status": "Merged 2025-10-16; public record checked 2026-10-06." } ], "learning": "Maintainer review challenged placement of logic, preserved behavior and whether new configuration was needed. One proposed configuration change was withdrawn after review.", "boundary": "These are specific contributions, not evidence of a total merged-PR count across organizations or employment at Graycore." }, "working_style": { "evidence": [ "Learns by building and investigating failures rather than treating a successful request as proof of system integrity.", "Uses AI tools for research, prototyping and implementation while retaining responsibility for decisions and validation.", "Open-beta runner feedback, maintainer review and co-founder collaboration inform the work.", "Practices explaining the same product differently to runners, collaborators and investors." ], "broader_context": [ "School: an ambitious Python personal-assistant attempt was left unfinished.", "First year: coding, DSA/competitive programming, video editing and storytelling; a startup founder commissioned a product video for a Y Combinator application.", "Third year: represents Runiverse in product conversations while continuing to develop technical and communication skills." ], "boundaries": "No client identity, fee, YC acceptance result, coding rank or standardized proficiency score is established." }, "role_comparison": { "useful_evidence": { "mobile_product_engineering": [ "Runiverse tracking/recovery", "Starbucks focus/keyboard/motion UX" ], "backend_reliability": [ "Runiverse activity identity and outbox", "UNI-X ownership integrity and job-payload investigation" ], "geospatial_engineering": [ "UNI-X H3 compaction/capture", "Custom dissolve and owner geometry" ], "ai_application_prototyping": [ "Div-AI retrieval/tool routing", "GuardFace research prototype with explicit evaluation limits" ], "rendering_and_frontend": [ "UNI-AI adaptive cards", "Astro portfolio and studio website stack" ], "collaboration_and_product": [ "Runiverse co-founder work", "Daffodil maintainer review", "Product storytelling" ] }, "method": "After the visitor supplies a role, map each important requirement to specific project evidence. Distinguish direct evidence, transferable experience and unknowns. Explain gaps and ask targeted technical/ownership questions; do not invent a hiring probability or claim every listed technology implies mastery.", "missing_information": [ "Formal employment history and dates", "Institution/degree/grades", "Team size or authorship beyond explicitly stated boundaries", "Current production scale, revenue and active users", "Availability and role preferences" ] } }