This repository is the source of truth for AJ's portfolio project and contains the conservative Astro static scaffold plus evidence-hardening work for the featured projects.
The portfolio positions AJ as:
- Primary identity: AI Systems Engineer
- Secondary identity: Product Engineer
The site is implemented as an evidence-safe static shell. This repo also contains the planning, architecture, design, content, risk, evidence, and handoff documentation that must guide future implementation.
The portfolio is deployed and production-verified on Vercel:
Live site: https://atishay-jain-portfolio.vercel.app
The current implementation is a production static portfolio backed by evidence-gated project data. Remaining project-proof gaps restrict only future claims and optional visuals; they are not release blockers for the published wording. Build, route, browser, responsive, accessibility, SEO, security-header, secret-scan, dependency-reachability, resume, and link checks are recorded in docs/37_PRODUCTION_DEPLOYMENT_REPORT.md.
Install locked dependencies:
npm ciRun the dev server:
npm run devRun validation:
npm test
ASTRO_TELEMETRY_DISABLED=1 npm run buildOn Windows PowerShell with script execution restrictions, use npm.cmd:
npm.cmd test
$env:ASTRO_TELEMETRY_DISABLED='1'; npm.cmd run build//projects/mahoraga/projects/hybrid-categorizer/projects/the-loop
Created:
- Astro, TypeScript, Tailwind, and MDX configuration.
- Data-driven homepage sections.
- Static featured project case-study shells.
- Evidence-safe pending visual panels for missing assets.
- Tests for featured ordering, verified links, missing evidence states, resume gating, and route definitions.
- Desktop and mobile screenshot QA under
docs/reviews/screenshots/. - Public CTAs for owner-approved GitHub, SDE resume, LinkedIn, and email destinations.
- Published primary SDE resume CTA at
/resume/atishay-jain-sde-resume.pdf. - Imported real candidate visuals under
public/assets/projects/...from verified external repos. - Hardened featured case-study pages with documented architecture, evidence status, limitations, missing proof, and next-evidence sections.
- Replaced Hybrid
predict.pngandmemory.pngwith safe sample-data frontend UI captures. - Added
docs/19_HYBRID_ROUTING_PROOF.mdand sample/unmeasured Hybrid routing rows without benchmark claims. - Added
docs/20_HYBRID_BENCHMARK_RESULTS.mdwith local read-only ONNX classifier/routing benchmark results. - Added
docs/21_HYBRID_HISTORY_AI_INSIGHTS_EVIDENCE.mddocumenting the click-through History AI Insight panel and the limits of the current list-view screenshot. - Conservative Signal Lab-aligned design-system polish for the static shell without changing evidence states or activating blocked CTAs.
- Fresh polished screenshot QA under
docs/reviews/screenshots/with review notes indocs/22_SCREENSHOT_QA_REVIEW.md. - Featured case-study copy polish for Mahoraga, Hybrid, and The Loop while preserving evidence gaps, local-only benchmark scope, and pending role/contribution details.
- Role/contribution evidence framework in
docs/23_ROLE_CONTRIBUTION_EVIDENCE.mdwith owner-reviewed role labels, contribution summaries, collaborative scope, still-to-verify items, and unsafe wording to avoid. - Public case-study contribution sections for Mahoraga, Hybrid GenAI Transaction Categorizer, and The Loop.
- Launch-readiness audit in
docs/24_LAUNCH_READINESS_AUDIT.mdwith a 72/100 score: conditionally ready for conservative soft launch, blocked for polished/final launch by visual/proof gaps. - The Loop live visual evidence update in
docs/25_THE_LOOP_VISUAL_EVIDENCE_UPDATE.mdwith owner-provided public-safe deployment screenshots and an updated 82/100 launch-readiness estimate.
Evidence still gated for future claims and optional media:
- Hybrid endpoint latency, Qwen fallback latency, cost, fallback-rate, and production claims until full endpoint/fallback benchmark evidence exists.
- Hybrid backend/model-output claims until the model/backend path is captured or benchmarked end to end.
- Hybrid History AI Insight visual proof until a safe sample transaction is opened and the real AI Insight panel is captured from
/transaction-insight. - Optional The Loop admin/chat/carpool visuals only if those flows will be visually featured; raw profile/friends screenshots remain unpublished and only redacted copies are public assets.
- Exact Mahoraga file/module ownership and exact The Loop module ownership unless supporting evidence or teammate confirmation is documented.
- Astro/Vitest major-version maintenance remains separate from the verified static release; current high advisories are not reachable in the deployed architecture.
Future agents should start here:
docs/00_READ_FIRST.mddocs/01_CURRENT_STATUS.mddocs/09_AI_HANDOFF.mdPROJECT_CONTEXT.mdPOSITIONING.mdFEATURED_PROJECT_SELECTION.mdPORTFOLIO_RISK_AUDIT.mddocs/10_PROJECT_EVIDENCE_REGISTRY.mddocs/11_LINK_REGISTRY.md
DESIGN.mdPROJECT_CONTEXT.mdFEATURED_PROJECT_SELECTION.mdPOSITIONING.mdCONTENT_ARCHITECTURE.mdPORTFOLIO_RISK_AUDIT.mdSITE_BLUEPRINT.mdDESIGN_SYSTEM.mdASSET_REQUIREMENTS.mddocs/00_READ_FIRST.mdthroughdocs/25_THE_LOOP_VISUAL_EVIDENCE_UPDATE.md
- Astro remains the default implementation stack.
- Static content comes before motion.
- Evidence comes before claims.
- Mahoraga, Hybrid GenAI Transaction Categorizer, and The Loop remain the featured projects.
- No fake screenshots, fake dashboards, fake metrics, or placeholder public links.
- Local Git repository exists.
- Remote:
https://github.com/Atishay9828/portfolio.git - GitHub repository:
https://github.com/Atishay9828/portfolio - Default branch:
master - Visibility: Public
- README exists and renders on GitHub.