Upload app.py with huggingface_hub

app.py

@@ -1,2 +1,507 @@
+import gradio as gr
+import pandas as pd
+import numpy as np
+import plotly.graph_objects as go
+from sklearn.feature_extraction.text import TfidfVectorizer
+from sklearn.cluster import KMeans
+from sklearn.preprocessing import normalize
+from sklearn.metrics.pairwise import cosine_similarity
+from sklearn.decomposition import PCA
+import ast, io
+from collections import Counter
 
-# paste the ENTIRE app.py code here after this line
+# ── Colors ────────────────────────────────────────
+COLORS5      = ["#6c63ff","#22d3ee","#f472b6","#fbbf24","#4ade80"]
+COLORS5_RGBA = ["rgba(108,99,255,0.2)","rgba(34,211,238,0.2)",
+                "rgba(244,114,182,0.2)","rgba(251,191,36,0.2)","rgba(74,222,128,0.2)"]
+ACCENT       = "#6c63ff"
+PLOTLY_THEME = dict(
+    paper_bgcolor="rgba(15,15,23,1)",
+    plot_bgcolor ="rgba(15,15,23,1)",
+    font=dict(family="sans-serif", color="#a8a6a0", size=12),
+    xaxis=dict(gridcolor="#1e1e2e", linecolor="#1e1e2e"),
+    yaxis=dict(gridcolor="#1e1e2e", linecolor="#1e1e2e"),
+    margin=dict(l=20, r=20, t=40, b=20),
+)
+
+def GAP_COLS(g):
+    if g > 0.89: return "#f87171"
+    if g > 0.79: return "#fbbf24"
+    return "#4ade80"
+
+def severity(g):
+    if g > 0.89: return "🔴 CRITICAL"
+    if g > 0.79: return "🟡 MODERATE"
+    return "🟢 LOW"
+
+def parse_skills(s):
+    try:
+        lst = ast.literal_eval(s)
+        return [x.strip().lower() for x in lst if x.strip()]
+    except:
+        return [x.strip().lower() for x in str(s).split(',') if x.strip()]
+
+# ── Demo data ─────────────────────────────────────
+def generate_demo_data():
+    rng = np.random.default_rng(42)
+    intern_pools = {
+        "Data Scientist":    ["python","machine learning","statistics","pandas","numpy","data analysis","regression","classification","visualization","jupyter"],
+        "Software Engineer": ["python","java","javascript","git","docker","api","algorithms","sql","testing","agile"],
+        "Business Analyst":  ["excel","powerpoint","project management","communication","stakeholder","reporting","business intelligence","ms office","presentation","analysis"],
+        "Finance Analyst":   ["excel","financial modeling","accounting","ledger","budget","forecasting","microsoft","audit","tax","compliance"],
+        "ML Engineer":       ["python","deep learning","tensorflow","pytorch","neural networks","nlp","machine learning","cuda","model deployment","feature engineering"],
+        "Web Developer":     ["html","css","javascript","react","git","responsive design","nodejs","rest api","typescript","webpack"],
+        "Data Analyst":      ["sql","tableau","python","excel","data cleaning","visualization","reporting","statistics","powerbi","etl"],
+        "DevOps Engineer":   ["docker","kubernetes","ci/cd","linux","aws","terraform","monitoring","git","bash","cloud"],
+    }
+    industry_pools = {
+        "Data Engineer":      ["sql","python","spark","aws","cloud","etl","data pipeline","airflow","kafka","dbt","azure","databricks"],
+        "ML Engineer":        ["python","aws","docker","kubernetes","mlflow","deep learning","cloud","machine learning","ci/cd","model serving"],
+        "Cloud Architect":    ["aws","azure","gcp","cloud","terraform","kubernetes","microservices","networking","security","cost optimization"],
+        "Data Scientist":     ["python","sql","machine learning","statistics","aws","communication","agile","data visualization","experiment design","causal inference"],
+        "Analytics Engineer": ["sql","dbt","python","analytics","data modeling","communication","business intelligence","airflow","testing","documentation"],
+    }
+    intern_rows, industry_rows = [], []
+    for role, skills in intern_pools.items():
+        n = rng.integers(80, 300)
+        for _ in range(n):
+            sample = rng.choice(skills, size=rng.integers(3,7), replace=False).tolist()
+            intern_rows.append({"Job_Role": role, "Intern_Skills": str(sample)})
+    for role, skills in industry_pools.items():
+        n = rng.integers(200, 800)
+        for _ in range(n):
+            sample = rng.choice(skills, size=rng.integers(4,9), replace=False).tolist()
+            industry_rows.append({"job_title": role, "job_skills": ", ".join(sample)})
+    return pd.DataFrame(intern_rows), pd.DataFrame(industry_rows)
+
+# ── Core pipeline ─────────────────────────────────
+def run_pipeline(intern_df, industry_df, k, top_n, min_df, max_df):
+    if 'Intern_Skills' not in intern_df.columns:
+        cols = [c for c in intern_df.columns if 'skill' in c.lower()]
+        intern_df['Intern_Skills'] = intern_df[cols[0]] if cols else ""
+    if 'Job_Role' not in intern_df.columns:
+        intern_df['Job_Role'] = intern_df.iloc[:,0]
+    if 'job_skills' not in industry_df.columns:
+        cols = [c for c in industry_df.columns if 'skill' in c.lower()]
+        industry_df['job_skills'] = industry_df[cols[0]] if cols else ""
+    if 'job_title' not in industry_df.columns:
+        industry_df['job_title'] = industry_df.iloc[:,0]
+
+    intern_df['skills_list']   = intern_df['Intern_Skills'].apply(parse_skills)
+    industry_df['skills_list'] = industry_df['job_skills'].apply(parse_skills)
+    intern_df   = intern_df[intern_df['skills_list'].map(len) > 0].copy()
+    industry_df = industry_df[industry_df['skills_list'].map(len) > 0].copy()
+    intern_df['skills_text']   = intern_df['skills_list'].apply(lambda x: ' '.join(x))
+    industry_df['skills_text'] = industry_df['skills_list'].apply(lambda x: ' '.join(x))
+
+    tfidf = TfidfVectorizer(stop_words='english', min_df=min_df, max_df=max_df)
+    tfidf.fit(pd.concat([intern_df['skills_text'], industry_df['skills_text']]))
+    features = tfidf.get_feature_names_out()
+
+    iv = normalize(tfidf.transform(intern_df['skills_text']))
+    jv = normalize(tfidf.transform(industry_df['skills_text']))
+
+    kmi = KMeans(n_clusters=k, random_state=42, n_init=10)
+    kmj = KMeans(n_clusters=k, random_state=42, n_init=10)
+    intern_df['Cluster']   = kmi.fit_predict(iv)
+    industry_df['Cluster'] = kmj.fit_predict(jv)
+
+    industry_mean = np.asarray(jv.mean(axis=0)).flatten()
+    sim = cosine_similarity(iv.toarray(), industry_mean.reshape(1,-1)).flatten()
+    intern_df['Similarity'] = sim
+    intern_df['Gap_Score']  = 1 - sim
+
+    top_idx             = industry_mean.argsort()[::-1][:top_n]
+    top_industry_skills = [features[i] for i in top_idx]
+    top_industry_set    = set(top_industry_skills)
+
+    profiles = {}
+    for cid in range(k):
+        center = kmi.cluster_centers_[cid]
+        top_intern = [features[i] for i in center.argsort()[::-1][:top_n]]
+        cdf = intern_df[intern_df['Cluster']==cid]
+        profiles[cid] = {
+            'label':      top_intern[0].title() if top_intern else f"Cluster {cid}",
+            'count':      len(cdf),
+            'avg_gap':    cdf['Gap_Score'].mean(),
+            'missing':    sorted(top_industry_set - set(top_intern))[:8],
+            'has':        sorted(top_industry_set & set(top_intern))[:6],
+            'top_skills': top_intern[:8],
+        }
+
+    heatmap_rows = []
+    for cid in range(k):
+        center = kmi.cluster_centers_[cid]
+        row = [float(center[tfidf.vocabulary_[s]]) if s in tfidf.vocabulary_ else 0 for s in top_industry_skills]
+        heatmap_rows.append(row)
+
+    pca = PCA(n_components=2, random_state=42)
+    sidx = np.random.choice(iv.shape[0], min(800, iv.shape[0]), replace=False)
+    coords = pca.fit_transform(iv[sidx].toarray())
+    df_pca = pd.DataFrame({
+        'x': coords[:,0], 'y': coords[:,1],
+        'cluster': intern_df.iloc[sidx]['Cluster'].values,
+        'gap':     intern_df.iloc[sidx]['Gap_Score'].values,
+        'role':    intern_df.iloc[sidx]['Job_Role'].values,
+    })
+
+    skill_freq = Counter([s for lst in intern_df['skills_list'] for s in lst]).most_common(20)
+    return intern_df, profiles, heatmap_rows, top_industry_skills, df_pca, skill_freq, k
+
+# ── Chart builders ────────────────────────────────
+def make_overview_charts(profiles, intern_df, K):
+    cids   = list(profiles.keys())
+    labels = [profiles[c]['label'] for c in cids]
+    gaps   = [profiles[c]['avg_gap'] for c in cids]
+
+    # Bar chart
+    fig_bar = go.Figure(go.Bar(
+        x=gaps, y=labels, orientation='h',
+        marker=dict(color=[GAP_COLS(g) for g in gaps], line=dict(width=0)),
+        text=[f"{g:.3f}" for g in gaps], textposition='outside',
+        textfont=dict(color="#a8a6a0", size=11),
+    ))
+    fig_bar.add_vline(x=0.89, line_dash="dash", line_color="#f87171", line_width=1,
+                      annotation_text="critical", annotation_font_color="#f87171")
+    fig_bar.add_vline(x=0.79, line_dash="dash", line_color="#fbbf24", line_width=1)
+    fig_bar.update_layout(**PLOTLY_THEME, height=300, title="Gap score by cluster",
+        xaxis=dict(range=[0.4,1.05], title="gap score", gridcolor="#1e1e2e"),
+        yaxis=dict(gridcolor="rgba(0,0,0,0)"), bargap=0.35)
+
+    # Donut
+    fig_pie = go.Figure(go.Pie(
+        labels=labels, values=[profiles[c]['count'] for c in cids],
+        hole=0.6, marker=dict(colors=COLORS5[:K], line=dict(color="#0a0a0f", width=2)),
+        textinfo='percent', textfont=dict(size=11),
+    ))
+    fig_pie.update_layout(**PLOTLY_THEME, height=300, title="Intern distribution",
+        showlegend=True, legend=dict(font=dict(size=10, color="#a8a6a0"), bgcolor="rgba(0,0,0,0)"))
+
+    # Violin
+    fig_vio = go.Figure()
+    for i, cid in enumerate(cids):
+        sub = intern_df[intern_df['Cluster']==cid]['Gap_Score']
+        fig_vio.add_trace(go.Violin(x=sub, name=profiles[cid]['label'],
+            line_color=COLORS5[i%len(COLORS5)],
+            fillcolor=COLORS5_RGBA[i%len(COLORS5_RGBA)],
+            opacity=0.8, box_visible=True, meanline_visible=True))
+    fig_vio.update_layout(**PLOTLY_THEME, height=250, title="Gap distribution",
+        xaxis_title="gap score", yaxis=dict(showgrid=False),
+        violingap=0.2, violinmode='overlay')
+
+    return fig_bar, fig_pie, fig_vio
+
+def make_heatmap(heatmap_rows, top_skills, profiles, K):
+    z = np.array(heatmap_rows)
+    cluster_labels = [f"C{c}: {profiles[c]['label']}" for c in range(K)]
+    fig = go.Figure(go.Heatmap(
+        z=z, x=top_skills, y=cluster_labels,
+        colorscale=[[0,"#1a0a2e"],[0.3,"#3c1a6b"],[0.6,"#6c3fc4"],[1.0,"#22d3ee"]],
+        text=[[f"{v:.3f}" for v in row] for row in z],
+        texttemplate="%{text}", textfont=dict(size=10, color="#e8e6e0"),
+        colorbar=dict(tickfont=dict(color="#a8a6a0"))
+    ))
+    fig.update_layout(**PLOTLY_THEME, height=350, title="Skill presence heatmap",
+        xaxis=dict(tickangle=-30, gridcolor="rgba(0,0,0,0)"),
+        yaxis=dict(gridcolor="rgba(0,0,0,0)"))
+    return fig
+
+def make_radar(heatmap_rows, top_skills, sel):
+    iv_r    = np.array(heatmap_rows[sel])
+    iv_norm = iv_r / iv_r.max() if iv_r.max() > 0 else iv_r
+    ind_r   = np.ones(len(top_skills)) * 0.9
+    cats    = top_skills + [top_skills[0]]
+    fig = go.Figure()
+    fig.add_trace(go.Scatterpolar(r=list(iv_norm)+[iv_norm[0]], theta=cats, fill='toself',
+        fillcolor="rgba(108,99,255,0.2)", line=dict(color=ACCENT, width=2), name="Intern"))
+    fig.add_trace(go.Scatterpolar(r=list(ind_r)+[ind_r[0]], theta=cats, fill='toself',
+        fillcolor="rgba(34,211,238,0.15)", line=dict(color="#22d3ee", width=2, dash="dash"),
+        name="Industry target"))
+    fig.update_layout(**PLOTLY_THEME, height=400, title="Radar — intern vs industry",
+        polar=dict(bgcolor="#0f0f17",
+            radialaxis=dict(visible=True, range=[0,1], gridcolor="#1e1e2e",
+                            tickfont=dict(color="#5a5a72", size=9)),
+            angularaxis=dict(gridcolor="#1e1e2e", tickfont=dict(color="#a8a6a0", size=10))),
+        legend=dict(font=dict(color="#a8a6a0"), bgcolor="rgba(0,0,0,0)"))
+    return fig
+
+def make_skill_charts(skill_freq, profiles, K):
+    skills_df = pd.DataFrame(skill_freq, columns=["skill","count"])
+    fig_sk = go.Figure(go.Bar(
+        x=skills_df['count'], y=skills_df['skill'], orientation='h',
+        marker=dict(color=skills_df['count'],
+                    colorscale=[[0,"#3c1a6b"],[1,"#22d3ee"]], line=dict(width=0)),
+        text=skills_df['count'], textposition='outside',
+        textfont=dict(color="#a8a6a0", size=10),
+    ))
+    fig_sk.update_layout(**PLOTLY_THEME, height=450, title="Top intern skills",
+        xaxis_title="frequency", yaxis=dict(gridcolor="rgba(0,0,0,0)"), bargap=0.3)
+
+    bub_x = [profiles[c]['avg_gap'] for c in range(K)]
+    bub_s = [profiles[c]['count']   for c in range(K)]
+    bub_l = [profiles[c]['label']   for c in range(K)]
+    fig_bub = go.Figure(go.Scatter(
+        x=bub_x, y=list(range(K)), mode='markers+text',
+        marker=dict(size=[s/max(bub_s)*60+20 for s in bub_s],
+                    color=[GAP_COLS(g) for g in bub_x], opacity=0.85,
+                    line=dict(color="#0a0a0f", width=2)),
+        text=bub_l, textposition="middle right",
+        textfont=dict(color="#a8a6a0", size=11),
+    ))
+    fig_bub.update_layout(**PLOTLY_THEME, height=300, title="Gap bubble chart",
+        xaxis=dict(title="avg gap score", range=[0.4,1.05]),
+        yaxis=dict(visible=False), showlegend=False)
+    return fig_sk, fig_bub
+
+def make_pca_charts(df_pca, intern_df, profiles, K):
+    fig_sc = go.Figure()
+    for cid in range(K):
+        sub = df_pca[df_pca['cluster']==cid]
+        fig_sc.add_trace(go.Scatter(
+            x=sub['x'], y=sub['y'], mode='markers',
+            name=f"C{cid}: {profiles[cid]['label']}",
+            marker=dict(color=COLORS5[cid%len(COLORS5)], size=5, opacity=0.7,
+                        line=dict(width=0)),
+            hovertemplate="<b>%{customdata[0]}</b><br>gap: %{customdata[1]:.3f}<extra></extra>",
+            customdata=np.stack([sub['role'], sub['gap']], axis=-1)
+        ))
+    fig_sc.update_layout(**PLOTLY_THEME, height=450, title="2D PCA cluster map",
+        xaxis_title="PC1", yaxis_title="PC2",
+        legend=dict(font=dict(color="#a8a6a0", size=10), bgcolor="rgba(0,0,0,0)"))
+
+    sample = intern_df.sample(min(500, len(intern_df)), random_state=42)
+    fig_gs = go.Figure()
+    for cid in range(K):
+        sub = sample[sample['Cluster']==cid]
+        fig_gs.add_trace(go.Scatter(
+            x=sub['Similarity'], y=sub['Gap_Score'], mode='markers',
+            name=f"C{cid}",
+            marker=dict(color=COLORS5[cid%len(COLORS5)], size=4, opacity=0.6,
+                        line=dict(width=0))
+        ))
+    fig_gs.update_layout(**PLOTLY_THEME, height=300, title="Gap score vs similarity",
+        xaxis_title="similarity", yaxis_title="gap score",
+        legend=dict(font=dict(color="#a8a6a0", size=10), bgcolor="rgba(0,0,0,0)"))
+    return fig_sc, fig_gs
+
+# ── Main analysis function ────────────────────────
+def analyze(intern_file, industry_file, k, top_n, min_df, max_df, cluster_sel, radar_sel):
+    # Load data
+    if intern_file is not None and industry_file is not None:
+        intern_df   = pd.read_csv(intern_file.name)
+        industry_df = pd.read_csv(industry_file.name)
+    else:
+        intern_df, industry_df = generate_demo_data()
+
+    k       = int(k)
+    top_n   = int(top_n)
+    min_df  = int(min_df)
+
+    intern_df, profiles, heatmap_rows, top_skills, df_pca, skill_freq, K = \
+        run_pipeline(intern_df, industry_df, k, top_n, min_df, max_df)
+
+    # KPI summary
+    total      = len(intern_df)
+    avg_gap    = intern_df['Gap_Score'].mean()
+    critical_n = sum(1 for p in profiles.values() if p['avg_gap'] > 0.89)
+
+    kpi_html = f"""
+    <div style="display:grid;grid-template-columns:repeat(4,1fr);gap:12px;margin:16px 0">
+      <div style="background:#13131e;border:1px solid #1e1e30;border-radius:12px;padding:1rem;text-align:center">
+        <div style="font-size:0.7rem;color:#5a5a72;text-transform:uppercase;letter-spacing:0.1em">Total interns</div>
+        <div style="font-size:1.8rem;font-weight:700;color:#f0ede6">{total:,}</div>
+      </div>
+      <div style="background:#13131e;border:1px solid #1e1e30;border-radius:12px;padding:1rem;text-align:center">
+        <div style="font-size:0.7rem;color:#5a5a72;text-transform:uppercase;letter-spacing:0.1em">Avg gap score</div>
+        <div style="font-size:1.8rem;font-weight:700;color:#f87171">{avg_gap:.3f}</div>
+      </div>
+      <div style="background:#13131e;border:1px solid #1e1e30;border-radius:12px;padding:1rem;text-align:center">
+        <div style="font-size:0.7rem;color:#5a5a72;text-transform:uppercase;letter-spacing:0.1em">Clusters</div>
+        <div style="font-size:1.8rem;font-weight:700;color:#f0ede6">{K}</div>
+      </div>
+      <div style="background:#13131e;border:1px solid #1e1e30;border-radius:12px;padding:1rem;text-align:center">
+        <div style="font-size:0.7rem;color:#5a5a72;text-transform:uppercase;letter-spacing:0.1em">Critical clusters</div>
+        <div style="font-size:1.8rem;font-weight:700;color:#f87171">{critical_n}/{K}</div>
+      </div>
+    </div>
+    """
+
+    # Cluster summary table
+    rows = ""
+    for cid, p in profiles.items():
+        sev = severity(p['avg_gap'])
+        rows += f"""<tr>
+          <td style="padding:8px 12px;color:#c8c6c0">Cluster {cid}</td>
+          <td style="padding:8px 12px;color:#a8a6a0">{p['label']}</td>
+          <td style="padding:8px 12px;color:#a8a6a0">{p['count']:,}</td>
+          <td style="padding:8px 12px;color:#f0ede6;font-weight:500">{p['avg_gap']:.3f}</td>
+          <td style="padding:8px 12px">{sev}</td>
+          <td style="padding:8px 12px;color:#f87171;font-size:0.85rem">{', '.join(p['missing'][:4])}</td>
+        </tr>"""
+
+    table_html = f"""
+    <div style="background:#0f0f17;border:1px solid #1e1e2e;border-radius:12px;overflow:hidden;margin:8px 0">
+      <table style="width:100%;border-collapse:collapse;font-size:0.88rem">
+        <thead>
+          <tr style="background:#13131e;border-bottom:1px solid #1e1e2e">
+            <th style="padding:10px 12px;color:#5a5a72;text-align:left;font-weight:400;font-size:0.75rem;text-transform:uppercase;letter-spacing:0.08em">Cluster</th>
+            <th style="padding:10px 12px;color:#5a5a72;text-align:left;font-weight:400;font-size:0.75rem;text-transform:uppercase;letter-spacing:0.08em">Name</th>
+            <th style="padding:10px 12px;color:#5a5a72;text-align:left;font-weight:400;font-size:0.75rem;text-transform:uppercase;letter-spacing:0.08em">Count</th>
+            <th style="padding:10px 12px;color:#5a5a72;text-align:left;font-weight:400;font-size:0.75rem;text-transform:uppercase;letter-spacing:0.08em">Avg gap</th>
+            <th style="padding:10px 12px;color:#5a5a72;text-align:left;font-weight:400;font-size:0.75rem;text-transform:uppercase;letter-spacing:0.08em">Severity</th>
+            <th style="padding:10px 12px;color:#5a5a72;text-align:left;font-weight:400;font-size:0.75rem;text-transform:uppercase;letter-spacing:0.08em">Top missing skills</th>
+          </tr>
+        </thead>
+        <tbody style="border-top:1px solid #1e1e2e">{rows}</tbody>
+      </table>
+    </div>"""
+
+    # Build all charts
+    fig_bar, fig_pie, fig_vio = make_overview_charts(profiles, intern_df, K)
+    fig_hm                    = make_heatmap(heatmap_rows, top_skills, profiles, K)
+
+    sel_r   = min(int(radar_sel) if str(radar_sel).isdigit() else 0, K-1)
+    fig_rad = make_radar(heatmap_rows, top_skills, sel_r)
+
+    fig_sk, fig_bub           = make_skill_charts(skill_freq, profiles, K)
+    fig_sc, fig_gs            = make_pca_charts(df_pca, intern_df, profiles, K)
+
+    # Recommendations HTML
+    training_plans = {
+        0: {"priority":["cloud platforms","agile methodologies","data analytics","SQL"],
+            "courses":["AWS Cloud Practitioner","Google Data Analytics","Agile Scrum","SQL Bootcamp"],
+            "timeline":"3–4 months"},
+        1: {"priority":["cloud deployment","agile workflows","database design","CI/CD"],
+            "courses":["AWS Solutions Architect","dbt Fundamentals","GitHub Actions","Airflow"],
+            "timeline":"2–3 months"},
+        2: {"priority":["data analytics","BI platforms","cloud tools","communication"],
+            "courses":["Power BI","Tableau","Cloud for Finance","Data Storytelling"],
+            "timeline":"3–4 months"},
+        3: {"priority":["agile/scrum","cloud services","data analysis","business context"],
+            "courses":["PSM I Scrum","AWS for Devs","Python for Data","Product Thinking"],
+            "timeline":"2–3 months"},
+        4: {"priority":["cloud ML","agile practices","analytics engineering","SQL"],
+            "courses":["MLflow + SageMaker","dbt Analytics","SQL for ML","Agile for AI"],
+            "timeline":"2 months"},
+    }
+
+    rec_html = ""
+    for cid in range(K):
+        p    = profiles[cid]
+        plan = training_plans.get(cid % 5, training_plans[0])
+        sev  = severity(p['avg_gap'])
+        pills = "".join(f'<span style="display:inline-block;padding:3px 10px;margin:2px;background:#2d1515;border:1px solid #5c2020;border-radius:6px;font-size:0.78rem;color:#f87171">{s}</span>' for s in plan['priority'])
+        courses = "".join(f'<span style="display:inline-block;padding:3px 10px;margin:2px;background:#1a1a2e;border:1px solid #2a2a42;border-radius:6px;font-size:0.78rem;color:#8888aa">{c}</span>' for c in plan['courses'])
+        rec_html += f"""
+        <div style="background:#13131e;border:1px solid #1e1e30;border-radius:12px;padding:1rem 1.2rem;margin-bottom:10px">
+          <div style="display:flex;justify-content:space-between;align-items:center;margin-bottom:8px">
+            <div style="font-weight:600;color:#f0ede6">Cluster {cid} — {p['label']}</div>
+            <div style="font-size:0.78rem;color:#5a5a72">{p['count']:,} interns · {plan['timeline']} · {sev}</div>
+          </div>
+          <div style="margin-bottom:6px"><span style="font-size:0.72rem;color:#5a5a72;text-transform:uppercase;letter-spacing:0.08em">Priority skills: </span>{pills}</div>
+          <div><span style="font-size:0.72rem;color:#5a5a72;text-transform:uppercase;letter-spacing:0.08em">Courses: </span>{courses}</div>
+        </div>"""
+
+    # Export CSV
+    export_df = intern_df[['Job_Role','Cluster','Gap_Score','Similarity']].copy()
+    export_df.to_csv("intern_gap_results.csv", index=False)
+
+    return (kpi_html, table_html,
+            fig_bar, fig_pie, fig_vio,
+            fig_hm, fig_rad,
+            fig_sk, fig_bub,
+            fig_sc, fig_gs,
+            rec_html,
+            "intern_gap_results.csv")
+
+# ── Gradio UI ─────────────────────────────────────
+css = """
+body { background: #0a0a0f !important; }
+.gradio-container { background: #0a0a0f !important; font-family: 'DM Sans', sans-serif; }
+.gr-panel { background: #0f0f17 !important; border: 1px solid #1e1e2e !important; }
+h1, h2, h3, label, .gr-block-label { color: #c8c6c0 !important; }
+.gr-button-primary { background: #6c63ff !important; border: none !important; color: white !important; }
+.gr-button-primary:hover { background: #7c73ff !important; }
+footer { display: none !important; }
+"""
+
+with gr.Blocks(css=css, title="◈ SkillScope — Intern Gap Analyzer") as demo:
+
+    gr.HTML("""
+    <div style="text-align:center;padding:2rem 0 1rem">
+      <div style="font-size:2.5rem;font-weight:800;color:#f0ede6;letter-spacing:-0.03em">◈ SkillScope</div>
+      <div style="font-size:0.85rem;color:#5a5a72;letter-spacing:0.15em;text-transform:uppercase;margin-top:4px">
+        Intern Skill Gap Analyzer · NLP + Clustering Intelligence
+      </div>
+    </div>
+    """)
+
+    # ── Controls ──────────────────────────────────
+    with gr.Row():
+        with gr.Column(scale=1):
+            gr.Markdown("### ⚙ Pipeline Config")
+            k_slider       = gr.Slider(2, 10, value=5, step=1, label="Clusters (K)")
+            topn_slider    = gr.Slider(5, 20, value=10, step=1, label="Top N skills")
+            mindf_slider   = gr.Slider(1, 10, value=2, step=1, label="TF-IDF min_df")
+            maxdf_slider   = gr.Slider(0.70, 1.0, value=0.95, step=0.05, label="TF-IDF max_df")
+            gr.Markdown("### 📂 Upload CSVs (optional)")
+            intern_file    = gr.File(label="Intern / Resume CSV", file_types=[".csv"])
+            industry_file  = gr.File(label="Industry / Jobs CSV", file_types=[".csv"])
+            gr.Markdown("### 🔍 Chart Options")
+            radar_sel      = gr.Number(value=0, label="Radar cluster index", precision=0)
+            cluster_sel    = gr.Number(value=0, label="Cluster select (unused)", precision=0, visible=False)
+            run_btn        = gr.Button("▶  Run Analysis", variant="primary")
+
+        # ── Results ───────────────────────────────
+        with gr.Column(scale=3):
+            with gr.Tabs():
+
+                with gr.Tab("📊 Overview"):
+                    kpi_out   = gr.HTML()
+                    table_out = gr.HTML()
+                    with gr.Row():
+                        bar_out = gr.Plot()
+                        pie_out = gr.Plot()
+                    vio_out   = gr.Plot()
+
+                with gr.Tab("🗺 Heatmap & Radar"):
+                    hm_out  = gr.Plot()
+                    rad_out = gr.Plot()
+
+                with gr.Tab("🔍 Skill Explorer"):
+                    sk_out  = gr.Plot()
+                    bub_out = gr.Plot()
+
+                with gr.Tab("🧭 2D Cluster Map"):
+                    sc_out = gr.Plot()
+                    gs_out = gr.Plot()
+
+                with gr.Tab("📋 Recommendations"):
+                    rec_out = gr.HTML()
+                    dl_out  = gr.File(label="Download intern results CSV")
+
+    run_btn.click(
+        fn=analyze,
+        inputs=[intern_file, industry_file, k_slider, topn_slider,
+                mindf_slider, maxdf_slider, cluster_sel, radar_sel],
+        outputs=[kpi_out, table_out,
+                 bar_out, pie_out, vio_out,
+                 hm_out, rad_out,
+                 sk_out, bub_out,
+                 sc_out, gs_out,
+                 rec_out, dl_out]
+    )
+
+    demo.load(
+        fn=analyze,
+        inputs=[intern_file, industry_file, k_slider, topn_slider,
+                mindf_slider, maxdf_slider, cluster_sel, radar_sel],
+        outputs=[kpi_out, table_out,
+                 bar_out, pie_out, vio_out,
+                 hm_out, rad_out,
+                 sk_out, bub_out,
+                 sc_out, gs_out,
+                 rec_out, dl_out]
+    )
+
+demo.launch()