/* global React, ReactDOM */ const { useState, useEffect, useRef, useMemo, useCallback } = React; // ── Hooks ─────────────────────────────────────────────────────────────────── function useTypewriter(words, pause = 1700, typingSpeed = 70, deletingSpeed = 35) { const [idx, setIdx] = useState(0); const [text, setText] = useState(""); const [phase, setPhase] = useState("typing"); useEffect(() => { const w = words[idx]; let t; if (phase === "typing") { if (text.length < w.length) t = setTimeout(() => setText(w.slice(0, text.length + 1)), typingSpeed); else t = setTimeout(() => setPhase("holding"), pause); } else if (phase === "holding") { t = setTimeout(() => setPhase("deleting"), pause); } else if (phase === "deleting") { if (text.length > 0) t = setTimeout(() => setText(w.slice(0, text.length - 1)), deletingSpeed); else { setPhase("typing"); setIdx((idx + 1) % words.length); } } return () => clearTimeout(t); }, [text, phase, idx, words, pause, typingSpeed, deletingSpeed]); return text; } function useReveal() { useEffect(() => { const els = document.querySelectorAll(".reveal"); const io = new IntersectionObserver( (entries) => entries.forEach((e) => { if (e.isIntersecting) e.target.classList.add("in"); }), { threshold: 0.12 } ); els.forEach((e) => io.observe(e)); return () => io.disconnect(); }); } function useActiveSection(ids) { const [active, setActive] = useState(ids[0]); useEffect(() => { function onScroll() { const y = window.scrollY + window.innerHeight * 0.35; let cur = ids[0]; for (const id of ids) { const el = document.getElementById(id); if (el && el.offsetTop <= y) cur = id; } setActive(cur); if (window.BG) window.BG.setSection(cur); } onScroll(); window.addEventListener("scroll", onScroll, { passive: true }); return () => window.removeEventListener("scroll", onScroll); }, [ids]); return active; } function normalizeLinkedInSync(sync) { if (!sync || !sync.meta || sync.meta.sync_status !== "ok") return null; if (!sync.profile || (!sync.profile.headline_short && !sync.profile.about_short)) return null; return { meta: sync.meta, profile: sync.profile, experience: Array.isArray(sync.experience) ? sync.experience.slice(0, 2) : [], featured: Array.isArray(sync.featured) ? sync.featured.slice(0, 2) : [], updates: Array.isArray(sync.updates) ? sync.updates.slice(0, 2) : [], }; } function mergeLinkedInIntoSite(baseData, sync) { const linkedin = normalizeLinkedInSync(sync); if (!linkedin) return baseData; const links = (baseData.links || []).map((link) => { if (link.label === "LinkedIn" && linkedin.profile.profile_url) { return { ...link, url: linkedin.profile.profile_url }; } return link; }); return { ...baseData, links, linkedin, }; } function formatLinkedInDate(value) { if (!value) return ""; const parsed = new Date(value); if (Number.isNaN(parsed.getTime())) return ""; return new Intl.DateTimeFormat("en-US", { month: "short", day: "numeric", year: "numeric" }).format(parsed); } function getLinkedInSourceView(meta) { const seeded = meta && meta.source === "linkedin-curated-seed"; const validatedDate = formatLinkedInDate(meta && meta.last_successful_sync_at); return { eyebrow: seeded ? "LinkedIn Snapshot" : "LinkedIn Signals", title: seeded ? "A curated LinkedIn snapshot, rendered safely while live sync catches up." : "Public profile changes, normalized before they touch the layout.", lead: seeded ? "This section is driven by a curated, validated snapshot that keeps the portfolio complete while public LinkedIn fetches are rate-limited." : "Machine-managed cards rendered only when public LinkedIn data passes fetch, parse, diff, and schema validation.", badge: seeded ? "Curated snapshot" : "Public sync", snapshotLabel: seeded ? "Profile snapshot" : "Current public snapshot", updatesLabel: seeded ? "Recent highlights" : "Latest public updates", experienceLabel: seeded ? "Experience in focus" : "Recent LinkedIn Experience", featuredLabel: seeded ? "Featured work" : "Featured from LinkedIn", note: meta && meta.warning ? meta.warning : seeded ? "Curated from portfolio-owned profile data and validated against the same schema as the live sync." : "Live public LinkedIn data is currently driving these cards.", metaLine: validatedDate ? `Last validated snapshot · ${validatedDate}` : seeded ? "Validated curated snapshot" : "Validated public snapshot", seeded, }; } // ── Atoms ─────────────────────────────────────────────────────────────────── function Tag({ children, accent }) { return {children}; } function Metric({ v, k, color, onMouseEnter, onMouseLeave }) { return (
{v} {k}
); } // ── TileCreature ──────────────────────────────────────────────────────────── // Tiny pseudo-3D SVG figures dropped into every content tile. Each `kind` // renders a self-contained scene (person + prop) with looping CSS animations // keyed off classNames defined in styles.css. `position` selects the corner // (br = bottom-right, default), `delay` desyncs neighbouring tiles so the // grid never animates in lockstep. function TileCreature({ kind, color, delay = 0, position = "br" }) { const styleVars = { "--c": color || "var(--accent)", "--d": (delay || 0) + "s" }; const cls = "creature creature--" + kind + " creature--" + position; let body = null; if (kind === "desk") { body = ( ); } else if (kind === "chart") { body = ( ); } else if (kind === "gear") { const teeth = [0, 45, 90, 135, 180, 225, 270, 315]; body = ( ); } else if (kind === "read") { body = ( ); } else if (kind === "paper") { body = ( ); } else if (kind === "build") { body = ( ); } else if (kind === "commit") { body = ( ); } else if (kind === "present") { body = ( ); } else if (kind === "teach") { body = ( ); } else { // wave (default) body = ( ); } return ( ); } function SectionTitle({ num, eyebrow, title, lead }) { return (

{eyebrow}

{title}

{num}
{lead &&

{lead}

}
); } function HeroConnectWorkflow({ data, onJump }) { const workflowLinks = [ { label: "Email", url: "mailto:" + data.profile.email, meta: "reply first", accent: true, }, { label: "LinkedIn", url: "https://www.linkedin.com/in/namit-shrivastava-baab47204/", meta: "network", }, { label: "GitHub", url: "https://github.com/namo507", meta: "code", }, { label: "Scholar", url: "https://scholar.google.com/citations?user=7bvTB-sAAAAJ&hl=en", meta: "papers", }, ]; return (

Connect workflow

Follow the signal, not the scavenger hunt.

Start with email, then branch into code, papers, or the professional internet.

Start here Connect survey + ML + applied systems
{workflowLinks.map((link, index) => ( {link.label} {link.meta} ))}
{data.profile.location} {data.profile.phone}
); } // ── Sections ──────────────────────────────────────────────────────────────── function Hero({ data, onJump }) { const word = useTypewriter(data.profile.typingWords); return (

{data.profile.eyebrow}

Currently building  {word}

{data.profile.headline}

{data.profile.summary}

GitHub ↗
{data.metrics.map((m, i) => { const metricFacts = [ "Geospatial integration across all 129,572 US census tracts at U. Michigan ISR — 100% broadband completeness achieved.", "Multi-source NLP pipeline: 1.1M+ Reddit posts + 40 NYT articles using DistilBERT. Presented at AAPOR 2025.", "Logo-similarity detection at Legistify — 2.4M images processed at 92% precision, cutting manual review by 70%.", "DistilBERT at 91.6% accuracy topped all 10 model comparisons on the EV sentiment analysis task.", ]; return ( window.BUDDY_SHOW?.({ title: m.value + " · " + m.label, fact: metricFacts[i] })} onMouseLeave={() => window.BUDDY_CLEAR?.()} /> ); })}
{/* Ribbon */}
{[...data.profile.typingWords, "Survey-aware AI", "Causal inference", "Geospatial epidemiology", "Privacy-preserving deployment", ...data.profile.typingWords].map((w, i) => ( ● {w} ))}
{/* interests cloud */}

Research interests

{data.interests.map((i) => {i})}
); } function CVSection({ data }) { return (
{/* Impact */}
{data.results.map((r, i) => { const c = ["#69d7c3","#8fb6ff","#f1b76a","#ff8a9b"][i]; return (
window.BUDDY_SHOW?.({ title: r.metric + " — " + r.title, fact: r.detail })} onMouseLeave={() => window.BUDDY_CLEAR?.()} >

Impact

{r.metric}

{r.title}

{r.detail}

); })}
{/* Experience timeline */}

Professional Experience

{data.experience.map((r, i) => { const c = ["#69d7c3","#8fb6ff","#f1b76a","#ff8a9b","#bcd86c","#c8a4ff"][i % 6]; return (
window.BUDDY_SHOW?.({ title: r.role + " @ " + r.org, fact: r.bullets[0] })} onMouseLeave={() => window.BUDDY_CLEAR?.()} >

{r.dates} · {r.location}

{r.role}

{r.org}

    {r.bullets.map((b, j) =>
  • {b}
    • )}
); })}
{/* Skills */}

Technical Skills

{data.skills.map((s, i) => { const c = ["#69d7c3","#8fb6ff","#f1b76a","#ff8a9b","#bcd86c","#c8a4ff","#7ad6ff"][i % 7]; return (
window.BUDDY_SHOW?.({ title: s.name + " · " + s.score + "%", fact: s.summary + " · Tools: " + s.tools.slice(0, 4).join(", ") })} onMouseLeave={() => window.BUDDY_CLEAR?.()} >
{s.tier}
{s.name}
{s.score}%

{s.summary}

{s.tools.map((t) => {t})}
); })}
{/* Education */}

Education

{data.education.map((e, i) => { const c = ["#8fb6ff","#f1b76a"][i]; return (
window.BUDDY_SHOW?.({ title: e.degree, fact: e.institution + " · " + e.dates + " · " + e.highlights.join(", ") })} onMouseLeave={() => window.BUDDY_CLEAR?.()} >

{e.dates} · {e.location}

{e.degree}

{e.institution}

{e.highlights.map((h) => {h})}
{e.coursework.map((c2) => {c2})}
); })}
{/* Achievements + Service */}

Achievements

    {data.achievements.map((a) =>
  • {a}
    • )}

Service & Leadership

    {data.service.map((s) =>
  • {s.role} · {s.org} ({s.dates})
    {s.note}
    • )}
Download PDF CV Email
); } function PublicationsSection({ data }) { return (
{data.publications.map((p, i) => { const c = ["#8fb6ff","#ff8a9b"][i]; return ( window.BUDDY_SHOW?.({ title: p.category + " · " + p.date, fact: p.excerpt })} onMouseLeave={() => window.BUDDY_CLEAR?.()} >

{p.category} · {p.date}

{p.title}

{p.venue}

{p.excerpt}

{p.stats.map((s) => (
{s.v}
{s.k}
))}

{p.citation}

Read paper ↗
 ); })}
Google Scholar ↗ ORCID ↗
); } function ProjectsSection({ data }) { const [filter, setFilter] = useState("All"); const tags = useMemo(() => { const set = new Set(["All"]); data.projects.forEach((p) => p.tags.forEach((t) => set.add(t))); return [...set].slice(0, 10); }, [data.projects]); const visible = filter === "All" ? data.projects : data.projects.filter((p) => p.tags.includes(filter)); const colors = ["#69d7c3","#8fb6ff","#f1b76a","#ff8a9b","#bcd86c","#c8a4ff","#7ad6ff"]; return (
{tags.map((t) => ( ))}
{visible.map((p, i) => { const c = colors[i % colors.length]; return ( window.BUDDY_SHOW?.({ title: "PROJECT " + p.id + " · " + p.type, fact: p.excerpt + " Tags: " + p.tags.join(", ") + "." })} onMouseLeave={() => window.BUDDY_CLEAR?.()} >

PROJECT · {p.id}

{p.title}

{p.type} · {p.date} · {p.venue}

{p.excerpt}

{p.tags.map((t) => {t})}
Open ↗
 ); })}
); } function GitHubSection({ data }) { const g = data.github; return (
{g.stats.map((s) => { const ghFacts = { "Public repositories": "41 public repos built across NLP, ML, survey methodology, and web development — all at github.com/namo507.", "Original builds": "28 repos built from scratch, not forks. Full ownership of design, code, and documentation.", "Starred spotlight": "7 repositories pinned as featured work: AAPOR EV Sentiment, Moneyball FC, office-doc-redactor, and more.", "Total accessible": "57 total accessible repositories including collaborations, class projects, and forks.", }; return ( window.BUDDY_SHOW?.({ title: s.v + " " + s.k, fact: ghFacts[s.k] || "GitHub @namo507 — 57 total repositories." })} onMouseLeave={() => window.BUDDY_CLEAR?.()} /> ); })}

Language Mix

{g.languageMix.map((l) => {l.lang} · {l.count})}

What this code covers

  • Research pipelines: survey methodology, NLP, multilevel modeling, public opinion analysis.
  • Applied AI prototypes: Office add-ins, n8n workflows, simulation frameworks.
  • Repository docs that link source code back to portfolio work and CV artifacts.

Featured Repositories

{g.featured.map((r, i) => { const c = ["#69d7c3","#8fb6ff","#f1b76a","#ff8a9b","#bcd86c","#c8a4ff","#7ad6ff"][i % 7]; return ( window.BUDDY_SHOW?.({ title: r.name + " · " + r.lang, fact: r.desc + (r.metric ? " · " + r.metricLabel + ": " + r.metric : "") })} onMouseLeave={() => window.BUDDY_CLEAR?.()} >

● {r.lang}

{r.name}

{r.desc}

{r.metric} {r.metricLabel}
 ); })}

Recently Updated

{g.recent.map((r) => (

● {r.lang}

{r.name}

{r.updated}

 ))}
@namo507 on GitHub ↗ All repositories Starred spotlight
); } function LinkedInSection({ data }) { const linkedin = data.linkedin; if (!linkedin) return null; const sourceView = getLinkedInSourceView(linkedin.meta); return (

{sourceView.snapshotLabel}

{sourceView.badge}

{linkedin.profile.headline_short}

{linkedin.profile.about_short}

{sourceView.note}

{(linkedin.profile.top_skills || []).slice(0, 6).map((skill) => {skill})}

{[linkedin.profile.current_role, linkedin.profile.organization, linkedin.profile.location].filter(Boolean).join(" · ")}

{sourceView.metaLine}

{sourceView.updatesLabel}

{linkedin.updates.length} cards
{linkedin.updates.length ? linkedin.updates.map((item) => ( {item.posted_relative || item.kind} {item.title}

{item.summary_short}

 )) : (

{sourceView.seeded ? "The curated snapshot is live, but no highlight cards were defined in the current seed data." : "Validated LinkedIn data is available, but the public profile did not expose recent activity cards on the last successful sync."}

)}
{linkedin.experience.length ? ( <>

{sourceView.experienceLabel}

{linkedin.experience.map((item) => (

{[item.date_range, item.location].filter(Boolean).join(" · ")}

{item.role}

{item.organization}

    {(item.bullets && item.bullets.length ? item.bullets : [item.description_short]).slice(0, 2).map((bullet) =>
  • {bullet}
    • )}
{item.canonical_url ? : null}
))}
) : null} {linkedin.featured.length ? ( <>

{sourceView.featuredLabel}

{linkedin.featured.map((item) => (

{[item.type, item.subtitle].filter(Boolean).join(" · ")}

{item.title}

{item.summary_short}

Open item ↗
 ))}
) : null}
); } function TalksSection({ data }) { return (
{data.talks.map((t, i) => ( window.BUDDY_SHOW?.({ title: t.type + " · " + t.date, fact: t.excerpt })} onMouseLeave={() => window.BUDDY_CLEAR?.()} >

{t.type} · {t.date} · {t.location}

{t.title}

{t.venue}

{t.excerpt}

Conference page ↗
 ))}

Snapshot

{data.talks.length} talk · 1 distinct venue · 2025

Most recent presentation: AAPOR 2025, St. Louis, MO. Looking ahead to AAPOR 2026 and methods workshops in 2026.

); } function TeachingSection({ data }) { return (
{data.teaching.map((t, i) => { const c = ["#bcd86c","#c8a4ff"][i]; return (
window.BUDDY_SHOW?.({ title: t.title + " · " + t.type, fact: t.excerpt + " · " + t.bullets.join(" · ") })} onMouseLeave={() => window.BUDDY_CLEAR?.()} >

{t.type} · {t.date}

{t.title}

{t.venue}

{t.excerpt}

    {t.bullets.map((b) =>
  • {b}
    • )}
); })}
); } function ContactSection({ data }) { const linkedinUrl = data.linkedin?.profile?.profile_url || "https://www.linkedin.com/in/namit-shrivastava-baab47204/"; return (

Connect

Bring survey puzzles, ML tangles, geospatial detours, or just a well-aimed note.

Email · {data.profile.email} LinkedIn ↗ GitHub ↗ Google Scholar ↗ ORCID ↗

{data.profile.location} · {data.profile.phone}

© 2026 Namit Shrivastava · Classic view · Source ↗

); } // ── AI Buddy helpers ───────────────────────────────────────────────────────── function getPrecomputedResponse(query, data) { const q = query.toLowerCase(); if (q.match(/publish|paper|article|research|journal|conference|scholar/)) { return "Namit has 2 research outputs: a Springer journal article on supply chain sustainability using fuzzy-AHP with 200+ experts (2024), and an AAPOR 2025 conference paper on EV sentiment analysis using DistilBERT at 91.6% accuracy over 1.1M+ posts."; } if (q.match(/skill|technolog|program|python|ml\b|machine learning|stack/)) { return "Top skills: Survey Methodology (98%), AI/ML (95%), Core AI Systems (91%), Programming (84%). Stack: Python, R, PyTorch, TensorFlow, Hugging Face, LangChain, and cloud platforms (AWS, Azure, GCP)."; } if (q.match(/educat|degree|university|college|gpa|grade|bits|umd/)) { return "M.S. in Survey & Data Science at UMD (GPA 3.814/4.0, JPSM Dean's Fellow 2025-26). B.E. in Civil Engineering with Data Science minor from BITS Pilani (GPA 3.327/4.0)."; } if (q.match(/project|build|portfolio|work/)) { return "20+ projects: EV sentiment (1.1M+ posts, 91.6% accuracy), football market value (7,023 rows, €38bn), LEXNet (97% smaller CNN), geospatial broadband analysis (129K+ tracts). See the Projects section!"; } if (q.match(/contact|email|reach|connect|message|phone/)) { return "Email: " + data.profile.email + " · LinkedIn: linkedin.com/in/namit-shrivastava-baab47204/ · GitHub: github.com/namo507. Based in " + data.profile.location + "."; } if (q.match(/github|repo|code|open.?source|commit/)) { return "41 public repos on GitHub (@namo507). Featured: AAPOR EV Sentiment, Project_Moneyball_FC, office-doc-redactor, live-meeting-copilot, career-ops. Python, R, TypeScript stack."; } if (q.match(/linkedin/)) { return "LinkedIn profile auto-synced via GitHub Actions every 5 days. Data passes fetch → parse → diff → schema validation before rendering into the portfolio."; } if (q.match(/ollama|local.?llm|llm|model|ai\b/)) { return "To unlock AI-powered responses, install Ollama at ollama.ai and run: \"ollama pull llama3.2\". Once running on localhost:11434, I'll connect automatically and give richer answers!"; } if (q.match(/fellowship|award|achiev|honor|recogni/)) { return "Awards: JPSM Dean's Fellowship (AY 2025-26), 1st rank in HRD among 180 students at BITS Pilani, Top 10 in Water & Wastewater Treatment, Microsoft Azure AI Fundamentals certified."; } if (q.match(/teach|course|student|ta\b|assistant|surv735/)) { return "TA for SURV735 (Data Privacy & Confidentiality) at JPSM, UMD — guiding 23 grad students. Also redesigned Canvas LMS for 10+ instructors, 125+ students: +30% satisfaction, -40% setup time."; } if (q.match(/experience|job|work history|role|position|internship/)) { return "5 roles: Social Data Science Center (UMD), U. Michigan ISR, JPSM Teaching TA, Legistify ML Engineer (92% precision, 2.4M images), Accenture (89% threat classification). Full timeline in CV section!"; } if (q.match(/survey|methodology|data.?collection|sampling|measurement/)) { return "Core research area: Survey & Data Science. Focus on trustworthy data integration, AI-enabled data collection quality assurance frameworks, causal inference, and privacy-preserving methods."; } if (q.match(/geospatial|census|broadband|michigan|epidemiol/)) { return "At U. Michigan ISR: deployed geospatial pipeline across 129,572 US census tracts & 3 RUCA strata. Achieved 100% broadband completeness, reduced missingness by 28.6%."; } if (q.match(/hello|hi\b|hey|greet|who are you|what can you do/)) { return "Hi! I'm Namit's research assistant. I know about his publications, projects, skills, and experience. Ask me anything — or hover over any tile for instant insights about that section!"; } return "Namit is a Graduate Researcher at UMD (GPA 3.814, Dean's Fellow) specializing in survey methodology, causal inference, and responsible AI. 20+ projects, 2 publications, 41 GitHub repos. What would you like to know?"; } // ── AiBuddy Component ──────────────────────────────────────────────────────── function AiBuddy({ data, activeSection }) { const [open, setOpen] = useState(false); const [messages, setMessages] = useState([{ role: "assistant", content: "Hi! I'm Namit's research assistant ✦ Hover any tile for instant insights, or ask me about his research, projects, and skills!", }]); const [input, setInput] = useState(""); const [insight, setInsight] = useState(null); const [ollamaStatus, setOllamaStatus] = useState("checking"); const [ollamaModel, setOllamaModel] = useState(null); const [thinking, setThinking] = useState(false); const [showTooltip, setShowTooltip] = useState(false); const [factIdx, setFactIdx] = useState(0); const msgEndRef = useRef(null); const inputRef = useRef(null); // Expose global setters so tile hover handlers (in sections) can feed insights useEffect(() => { window.BUDDY_SHOW = (ins) => { setInsight(ins); setShowTooltip(true); }; window.BUDDY_CLEAR = () => { setInsight(null); setShowTooltip(false); }; return () => { delete window.BUDDY_SHOW; delete window.BUDDY_CLEAR; }; }, []); // Check Ollama availability on mount (silent — never blocks UI) useEffect(() => { (async () => { try { const ctrl = new AbortController(); const tid = setTimeout(() => ctrl.abort(), 2800); const res = await fetch("http://localhost:11434/api/tags", { signal: ctrl.signal }); clearTimeout(tid); if (!res.ok) { setOllamaStatus("unavailable"); return; } const d = await res.json(); const models = d.models || []; if (models.length > 0) { setOllamaStatus("available"); const preferred = ["llama3.2", "llama3", "mistral", "phi3", "qwen", "gemma"]; const hit = preferred.reduce((f, p) => f || models.find(m => m.name.startsWith(p)), null); setOllamaModel(hit ? hit.name : models[0].name); } else { setOllamaStatus("unavailable"); } } catch { setOllamaStatus("unavailable"); } })(); }, []); // Auto-scroll useEffect(() => { if (open && msgEndRef.current) msgEndRef.current.scrollIntoView({ behavior: "smooth" }); }, [messages, open]); // Focus input on open useEffect(() => { if (open && inputRef.current) setTimeout(() => inputRef.current?.focus(), 180); }, [open]); const sectionInsights = useMemo(() => ({ home: { title: "Portfolio Overview", fact: "129K+ census tracts · 1.1M+ social posts processed · 2.4M images in ML workflows · 91.6% best model accuracy." }, cv: { title: "CV Highlights", fact: "GPA 3.814/4.0 at UMD · JPSM Dean's Fellow 2025-26 · 5 professional roles spanning research, ML engineering, and TA positions." }, publications: { title: "Research Output", fact: "Springer journal (2024) + AAPOR 2025 conference paper. DistilBERT 91.6% on 1.1M+ EV posts. +0.57 LLM sentiment bias discovered." }, projects: { title: "Project Portfolio", fact: "20 projects: LEXNet (97% smaller CNN), Voice recognition (98.3%), Football market €38bn, Broadband pipeline 129K tracts." }, github: { title: "Code Portfolio", fact: "41 public repos · Python, R, TypeScript · Featured: AAPOR EV Analysis, Moneyball FC, office-doc-redactor, live-meeting-copilot." }, "linkedin-signals": { title: "LinkedIn Signals", fact: "Auto-synced via GitHub Actions every 5 days. Data passes fetch → parse → diff → schema validation before rendering." }, talks: { title: "Conference Talks", fact: "AAPOR 80th Annual, St. Louis MO, May 2025. Pipeline: 1.1M+ posts → DistilBERT → 10 Groq LLM variants compared." }, teaching: { title: "Teaching Practice", fact: "TA for SURV735 at JPSM. Canvas LMS redesign for 125+ students. +30% satisfaction · -40% setup time · 100% accessibility." }, contact: { title: "Contact", fact: "College Park, MD · namit507@gmail.com · Open to research collaborations, data science work, and survey methodology discussions." }, }), []); const funFacts = useMemo(() => [ { title: "📊 Census Scale", fact: "129,572 US census tracts analyzed — every single tract in the country — for broadband completeness research." }, { title: "🤖 Model Accuracy", fact: "DistilBERT hits 91.6% accuracy on 1.1M+ EV sentiment posts from Reddit & NYT at AAPOR 2025." }, { title: "⚡ LEXNet Speed", fact: "LEXNet CNN: 97% smaller, 93% faster than baseline, +4% accuracy improvement over standard architectures." }, { title: "🎙️ Voice AI", fact: "Voice gender recognition: 98.3% accuracy — only 11 misclassifications on the full held-out test set." }, { title: "⚽ Football Markets", fact: "€38bn football transfer market modeled with 7,023 player-season observations. 78% of player-level variance explained." }, { title: "🩸 Community Impact", fact: "Organized a 2-day blood donation drive: 60+ volunteers, 1,000+ donor records, 844 successful donations." }, { title: "🏆 Dean's Fellow", fact: "JPSM Dean's Fellowship awarded at University of Maryland for AY 2025-26." }, { title: "🔍 Sentiment Bias", fact: "+0.57 systematic positive bias found in LLM sentiment outputs vs Reddit data. F(2,549)=28.43, p<0.001." }, { title: "📡 Broadband Pipeline", fact: "Geospatial pipeline: 129,572 US census tracts × 3 RUCA strata — 100% completeness, -28.6% baseline missingness." }, { title: "🖼️ Image ML", fact: "Multilabel CNN on 42,520 road-surface images → 86.67% accuracy, 88.19% F1 across 5 distress categories." }, ], []); const displayInsight = insight || sectionInsights[activeSection] || null; const buildSystemPrompt = useCallback(() => ( "You are an AI portfolio assistant for " + data.profile.name + ", a " + data.profile.role + " at UMD. Help visitors learn about their work concisely (max 2-3 sentences). " + "Key facts: GPA 3.814, JPSM Dean's Fellow 2025-26. " + "Publications: Springer supply chain article + AAPOR 2025 EV sentiment (91.6% DistilBERT). " + "Projects: LEXNet, football market analysis, geospatial broadband pipeline. " + "Email: " + data.profile.email ), [data]); const sendMessage = useCallback(async () => { const trimmed = input.trim(); if (!trimmed || thinking) return; setInput(""); const newMsgs = [...messages, { role: "user", content: trimmed }]; setMessages(newMsgs); setThinking(true); let reply; if (ollamaStatus === "available" && ollamaModel) { try { const ctrl = new AbortController(); const tid = setTimeout(() => ctrl.abort(), 35000); const res = await fetch("http://localhost:11434/api/chat", { method: "POST", headers: { "Content-Type": "application/json" }, body: JSON.stringify({ model: ollamaModel, messages: [ { role: "system", content: buildSystemPrompt() }, ...newMsgs.slice(-6).map(m => ({ role: m.role, content: m.content })), ], stream: false, }), signal: ctrl.signal, }); clearTimeout(tid); reply = res.ok ? ((await res.json()).message?.content || getPrecomputedResponse(trimmed, data)) : getPrecomputedResponse(trimmed, data); } catch { reply = getPrecomputedResponse(trimmed, data); } } else { reply = getPrecomputedResponse(trimmed, data); } setMessages(prev => [...prev, { role: "assistant", content: reply }]); setThinking(false); }, [input, messages, thinking, ollamaStatus, ollamaModel, data, buildSystemPrompt]); const handleKey = useCallback((e) => { if (e.key === "Enter" && !e.shiftKey) { e.preventDefault(); sendMessage(); } }, [sendMessage]); const ollamaLabel = ollamaStatus === "checking" ? "Checking Ollama…" : ollamaStatus === "available" ? "✓ Ollama · " + ollamaModel : "Ollama offline · keyword mode"; return ( <>
{showTooltip && displayInsight && !open && (
{displayInsight.title}
{displayInsight.fact}
)}
{open && (
✦ Research Assistant
{ollamaLabel}
{/* Rotating fun fact — click to cycle */}
setFactIdx(i => (i + 1) % funFacts.length)} title="Click for next fact"> ✦ Research fact · tap to rotate {funFacts[factIdx].title} — {funFacts[factIdx].fact}
{/* Current section or tile insight */} {displayInsight && (
{displayInsight.title} {displayInsight.fact}
)}
{messages.map((m, i) => (
{m.content}
))} {thinking && (
)}
setInput(e.target.value)} onKeyDown={handleKey} placeholder="Ask about research, projects…" disabled={thinking} aria-label="Ask a question" />
)} ); } // ── Top-level App ─────────────────────────────────────────────────────────── function App() { const data = useMemo(() => mergeLinkedInIntoSite(window.SITE, window.LINKEDIN_SYNC), []); const navIds = data.navigation.map((n) => n.id); const active = useActiveSection(navIds); useReveal(); const jump = useCallback((id) => { const el = document.getElementById(id); if (el) window.scrollTo({ top: el.offsetTop - 60, behavior: "smooth" }); }, []); return ( <> ); } const mountApp = () => { const root = ReactDOM.createRoot(document.getElementById("app")); root.render(); }; const linkedInSyncReady = window.LINKEDIN_SYNC_READY; if (linkedInSyncReady && typeof linkedInSyncReady.then === "function") { linkedInSyncReady.finally(mountApp); } else { mountApp(); }