Sync high-value al-folio features (theme/system, publications, charts, search, Docker)

.dockerignore

@@ -0,0 +1,9 @@
+_site
+.git
+.jekyll-cache
+.jekyll-metadata
+.sass-cache
+.bundle
+vendor
+Gemfile.lock
+.DS_Store

.github/workflows/update-citations.yml

@@ -0,0 +1,37 @@
+name: Update Google Scholar citations
+
+on:
+  schedule:
+    - cron: "0 3 * * 1" # weekly, Monday 03:00 UTC
+  workflow_dispatch: # allow manual runs
+
+permissions:
+  contents: write
+
+jobs:
+  update-citations:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.11"
+
+      - name: Install dependencies
+        run: pip install -r requirements.txt
+
+      - name: Fetch citation counts
+        run: python bin/update_scholar_citations.py
+
+      - name: Commit updated citations
+        run: |
+          git config user.name "github-actions[bot]"
+          git config user.email "github-actions[bot]@users.noreply.github.com"
+          git add _data/citations.yml
+          if git diff --staged --quiet; then
+            echo "No citation changes."
+          else
+            git commit -m "chore: update Google Scholar citations"
+            git push
+          fi

Dockerfile

@@ -0,0 +1,21 @@
+# Development image for building the al-folio site on a pinned Ruby.
+#
+# Rationale: newer host Rubies (e.g. Ruby 4.0) dropped several libraries from
+# the default gem set (ostruct, bigdecimal, ...), which breaks the legacy plugin
+# chain. Pinning Ruby 3.2 here gives a reliable local build/serve regardless of
+# the host toolchain. Deployment still happens via GitHub Actions.
+FROM ruby:3.2
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+      build-essential \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /srv/jekyll
+
+COPY Gemfile ./
+RUN gem install bundler && bundle install
+
+EXPOSE 4000 35729
+
+CMD ["bundle", "exec", "jekyll", "serve", \
+     "--host", "0.0.0.0", "--livereload", "--force_polling", "--watch"]

Gemfile

@@ -1,9 +1,11 @@
 source 'https://rubygems.org'
+
 group :jekyll_plugins do
     gem 'jekyll'
     gem 'jekyll-email-protect'
     gem 'jekyll-feed'
     gem 'jekyll-github-metadata'
+    gem 'jekyll-jupyter-notebook'
     gem 'jekyll-paginate-v2'
     gem 'jekyll-scholar'
     gem 'jekyll-twitter-plugin'
@@ -14,3 +16,13 @@ group :jekyll_plugins do
     gem 'unicode_utils'
     gem 'webrick'
 end
+
+# Ruby 4.0 removed these from the default gem set. Listing them explicitly keeps
+# the build working on Ruby 4.x while remaining harmless on Ruby 3.2 (where they
+# are still available as gems). The Docker image pins Ruby 3.2, so this is only
+# load-bearing for local builds on a newer host Ruby.
+gem 'ostruct'
+gem 'bigdecimal'
+gem 'csv'
+gem 'logger'
+gem 'faraday-retry'

_bibliography/papers.bib

@@ -30,6 +30,7 @@ @inproceedings{GohLWSB21Understanding
   address   = {Virtual Event / Milan, Italy},
   html      = {https://doi.org/10.1109/ICPR48806.2021.9413242},
   doi       = {10.1109/ICPR48806.2021.9413242},
+  google_scholar_id = {j3f4tGmQtD8C},
   arxiv     = {2004.10484},
   abbr      = {ICPR},
   code      = {https://github.com/garygsw/smooth-taylor},
@@ -59,6 +60,7 @@ @inproceedings{GohKZ18Twitter
   address    = {Singapore},
   html       = {https://doi.org/10.1109/ICDMW.2018.00097},
   doi        = {10.1109/ICDMW.2018.00097},
+  google_scholar_id = {4JMBOYKVnBMC},
   abbr       = {ICDM},
   code       = {https://github.com/garygsw/twitter-crowd-flow-prediction},
   slides     = {icdm2018/ICDMW2018_TCFP_Slides.pdf},
@@ -84,6 +86,7 @@ @inproceedings{OngGX15Sparsity
   address   = {Santa Clara, CA, USA},
   html      = {https://doi.org/10.1109/BigData.2015.7363995},
   doi       = {10.1109/BigData.2015.7363995},
+  google_scholar_id = {d1gkVwhDpl0C},
   abbr      = {IEEE BigData},
   pdf       = {ieeebigdata2015/IEEEBigData_SAIG_Paper.pdf}
 }
@@ -105,6 +108,7 @@ @inproceedings{YangGJZA15Forecast
   address   = {Santa Clara, CA, USA},
   html      = {https://doi.org/10.1109/BigData.2015.7363994},
   doi       = {10.1109/BigData.2015.7363994},
+  google_scholar_id = {u-x6o8ySG0sC},
   abbr      = {IEEE BigData},
   pdf       = {ieeebigdata2015/IEEEBigData_ForecastFMCG-II_Paper.pdf}
 }
@@ -126,6 +130,7 @@ @inproceedings{YangGXZA15Forecast
   address   = {Santa Clara, CA, USA},
   html      = {https://doi.org/10.1109/BigData.2015.7363993},
   doi       = {10.1109/BigData.2015.7363993},
+  google_scholar_id = {u5HHmVD_uO8C},
   abbr      = {IEEE BigData},
   pdf       = {ieeebigdata2015/IEEEBigData_ForecastFMCG-I_Paper.pdf}
 }
@@ -159,6 +164,7 @@ @inproceedings{YangGJZ16Spatial
   address   = {Washington DC, USA},
   html      = {https://doi.org/10.1109/BigData.2016.7841052},
   doi       = {10.1109/BigData.2016.7841052},
+  google_scholar_id = {RHpTSmoSYBkC},
   abbr      = {IEEE BigData},
   pdf       = {ieeebigdata2016/IEEEBigData_SpatialQuadTree_Paper.pdf}
 }
@@ -192,6 +198,7 @@ @inproceedings{YangGJZ16aForecast
   address   = {Washington DC, USA},
   html      = {https://doi.org/10.1109/BigData.2016.7841053},
   doi       = {10.1109/BigData.2016.7841053},
+  google_scholar_id = {_Qo2XoVZTnwC},
   abbr      = {IEEE BigData},
   pdf       = {ieeebigdata2015/IEEEBigData_ForecastFMCG-III_Paper.pdf}
 }
@@ -224,6 +231,7 @@ @inproceedings{GohAZ16Optimizing
   address   = {Washington DC, USA},
   html      = {https://doi.org/10.1109/BigData.2016.7841042},
   doi       = {10.1109/BigData.2016.7841042},
+  google_scholar_id = {e5wmG9Sq2KIC},
   abbr      = {IEEE BigData},
   pdf       = {ieeebigdata2016/IEEEBigData_SentimentDOE_Paper.pdf}
 }

_config.yml

@@ -219,6 +219,10 @@ enable_navbar_social:
   # navbar on the about page
 enable_project_categories: true # enables categorization of projects into # multiple categories
 enable_medium_zoom: true # enables image zoom feature (as on medium.com)
+enable_search: false # enables the ninja-keys command-palette site search
+# Google Scholar citation badge on publications (count sourced from _data/citations.yml)
+enable_publication_badges:
+  google_scholar: true
 
 # -----------------------------------------------------------------------------
 # Library versions

_data/citations.yml

@@ -0,0 +1,38 @@
+# Google Scholar citation counts, keyed by "<scholar_userid>:<google_scholar_id>".
+# Auto-refreshed by .github/workflows/update-citations.yml (bin/update_scholar_citations.py).
+# Values below were seeded from the Google Scholar profile (user SuI63bsAAAAJ).
+metadata:
+  last_updated: "2026-07-11"
+papers:
+  "SuI63bsAAAAJ:j3f4tGmQtD8C":
+    title: Understanding Integrated Gradients with SmoothTaylor for Deep Neural Network Attribution
+    year: 2021
+    citations: 63
+  "SuI63bsAAAAJ:u-x6o8ySG0sC":
+    title: "Forecast UPC-level FMCG demand, Part II: Hierarchical Reconciliation"
+    year: 2015
+    citations: 19
+  "SuI63bsAAAAJ:4JMBOYKVnBMC":
+    title: Twitter-Informed Crowd Flow Prediction
+    year: 2018
+    citations: 16
+  "SuI63bsAAAAJ:d1gkVwhDpl0C":
+    title: Sparsity Adjusted Information Gain for Feature Selection in Sentiment Analysis
+    year: 2015
+    citations: 14
+  "SuI63bsAAAAJ:_Qo2XoVZTnwC":
+    title: "Forecast UPC-level FMCG demand, Part III: Grouped Reconciliation"
+    year: 2016
+    citations: 9
+  "SuI63bsAAAAJ:u5HHmVD_uO8C":
+    title: "Forecast UPC-level FMCG demand, Part I: Exploratory Analysis and Visualization"
+    year: 2015
+    citations: 9
+  "SuI63bsAAAAJ:RHpTSmoSYBkC":
+    title: "Spatial Data Dimension Reduction using Quadtree: A Case Study on Satellite-derived Solar Radiation"
+    year: 2016
+    citations: 5
+  "SuI63bsAAAAJ:e5wmG9Sq2KIC":
+    title: Optimizing Performance of Sentiment Analysis through Design of Experiments
+    year: 2016
+    citations: 3

_data/cv.yml

@@ -24,12 +24,14 @@
       end_date: Present
       location: Singapore
       description:
-        - Work across the full data function end to end — data engineering, analytics, data science, and AI transformation — supporting management decision-making and fundraising
-        - Architected <b>Data Warehouse 2.0</b> on <a href="https://www.databricks.com"><b>Databricks</b></a>, migrating to a governed native catalog with CDC streaming via <b>AWS DMS</b> and cutting data-platform costs by ~80%
-        - Designed a company-wide AI operating model &mdash; a git-based knowledge base plus a governed <b>MCP</b> layer connecting AI agents to live company data (Databricks, Jira, GitHub, Drive, Gmail, Mixpanel) &mdash; and rolled out Claude and GPT/Codex org-wide with read-only and PII-ringfencing guardrails
-        - Built a self-healing LLM pipeline that auto-classifies Databricks job failures, self-repairs or opens PRs, and closes Jira tickets, eliminating manual triage
-        - Deployed predictive models, including quantile-regression doctor-demand forecasting, an AI Fraud/Waste/Abuse detection engine across 14 risk categories, and B2B corporate segmentation
-        - Data lead for the Series C fundraise data room; delivered self-serve natural-language analytics via <a href="https://www.databricks.com/product/ai-bi/genie"><b>Databricks Genie</b></a> and an MCP-connected agent
+        - "<b>Business Impact</b>: Drove regional data science initiatives (SG/MY/PH/HK/ID), delivering 10&ndash;30% incremental revenue and operational cost savings."
+        - "<b>Data Architecture</b>: Architected <b>Data Warehouse 2.0</b> on <a href='https://www.databricks.com'>Databricks</a>; migrated to a governed native catalog with CDC streaming via <b>AWS DMS</b>, slashing data platform costs by 80%."
+        - "<b>Company-Wide AI Transformation</b>: Designed a git-based knowledge base and a governed <b>MCP</b> layer connecting AI agents to live company data (Databricks, Jira, GitHub, Drive, Gmail, Mixpanel)."
+        - "<b>AI Governance</b>: Implemented read-only guardrails, PII ringfencing, and infrastructure guardrails while rolling out Claude and GPT/Codex org-wide."
+        - "<b>AI Agent Automation</b>: Built a self-healing LLM pipeline that auto-classifies Databricks job failures, self-repairs/opens PRs, and manages Jira tickets, eliminating manual triage."
+        - "<b>Clinical GenAI</b>: Deployed patient-facing &ldquo;AI Triage&rdquo; and doctor-facing &ldquo;Clinical Support&rdquo; bots using LangChain/LangGraph on AWS AgentCore, validated via in-house eval datasets."
+        - "<b>Self-Serve Analytics</b>: Integrated Databricks Genie with an MCP-connected agent for natural-language querying, reducing ad-hoc data requests from non-technical teams."
+        - "<b>Predictive Modeling</b>: Deployed quantile regression for doctor demand forecasting, an AI Fraud/Waste/Abuse detection engine across 14 risk categories, and B2B corporate segmentation models."
     - title: Data Scientist
       subtitle: "<i><a href='https://www.sea.com/home'>Sea Ltd</a> &mdash; <a href='https://www.garena.sg'>Garena</a></i>"
       year: 2021-2022

_data/venues.yml

@@ -14,4 +14,4 @@ ICDM:
     url: http://icdm.bigke.org
 
 SUTD:
-    url: https://www.sutd.edu.sg/
+    url: https://www.sutd.edu.sg/

_includes/header.html

@@ -76,11 +76,19 @@
           {% endif %}
           {% endif %}
           {% endfor %}
+          {% if site.enable_search %}
+            <div class = "toggle-container">
+              <a id = "search-toggle" role="button" title="Search (⌘K / Ctrl-K)">
+                  <i class="fas fa-search"></i>
+              </a>
+            </div>
+          {% endif %}
           {% if site.enable_darkmode %}
             <div class = "toggle-container">
-              <a id = "light-toggle">
-                  <i class="fas fa-moon"></i>
-                  <i class="fas fa-sun"></i>
+              <a id = "light-toggle" role="button" title="Cycle light / dark / system theme">
+                  <i class="fas fa-sun" title="Switch to dark theme"></i>
+                  <i class="fas fa-moon" title="Switch to system theme"></i>
+                  <i class="fas fa-adjust" title="Switch to light theme"></i>
               </a>
             </div>
           {% endif %}

_includes/jupyter_notebook.html

@@ -0,0 +1,18 @@
+{%- comment -%}
+  Embed a Jupyter notebook.
+
+  Preferred (needs Jupyter installed + jekyll-jupyter-notebook, which is in the Gemfile):
+      {% jupyter_notebook "assets/jupyter/blog.ipynb" %}
+
+  Portable fallback (no local Jupyter toolchain) — embed pre-rendered notebook HTML
+  or an nbviewer page in an iframe:
+      {% include jupyter_notebook.html path="/assets/jupyter/demo.html" %}
+      {% include jupyter_notebook.html url="https://nbviewer.org/github/<user>/<repo>/blob/main/nb.ipynb" %}
+{%- endcomment -%}
+{%- if include.path -%}
+  {%- assign nb_src = include.path | relative_url -%}
+{%- else -%}
+  {%- assign nb_src = include.url -%}
+{%- endif -%}
+<iframe class="jupyter-notebook-iframe" src="{{ nb_src }}" loading="lazy"
+        title="{{ include.title | default: 'Jupyter notebook' }}"></iframe>

_includes/scripts/diagrams.html

@@ -0,0 +1,44 @@
+<!--
+  Rich-content renderers, loaded only on pages that opt in via front matter:
+    chart_js: true    -> Chart.js          (<canvas> + new Chart(...))
+    plotly: true      -> Plotly.js         (<div> + Plotly.newPlot(...))
+    mermaid: true     -> Mermaid diagrams   (<div class="mermaid">...)
+    pseudocode: true  -> pseudocode.js      (<pre class="pseudocode">...)  [needs KaTeX]
+    tikzjax: true     -> TikZJax            (<script type="text/tikz">...)
+-->
+{% if page.chart_js %}
+<script src="https://cdn.jsdelivr.net/npm/chart.js@4"></script>
+{% endif %}
+
+{% if page.plotly %}
+<script src="https://cdn.plot.ly/plotly-2.35.2.min.js" charset="utf-8"></script>
+{% endif %}
+
+{% if page.mermaid %}
+<script src="https://cdn.jsdelivr.net/npm/mermaid@11/dist/mermaid.min.js"></script>
+<script>
+  mermaid.initialize({
+    startOnLoad: true,
+    theme: document.documentElement.getAttribute("data-theme") === "dark" ? "dark" : "default",
+  });
+</script>
+{% endif %}
+
+{% if page.pseudocode %}
+<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/katex.min.css">
+<script defer src="https://cdn.jsdelivr.net/npm/[email protected]/dist/katex.min.js"></script>
+<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/[email protected]/build/pseudocode.min.css">
+<script defer src="https://cdn.jsdelivr.net/npm/[email protected]/build/pseudocode.min.js"></script>
+<script>
+  window.addEventListener("load", function () {
+    if (window.pseudocode) {
+      pseudocode.renderClass("pseudocode", { lineNumber: true });
+    }
+  });
+</script>
+{% endif %}
+
+{% if page.tikzjax %}
+<link rel="stylesheet" href="https://tikzjax.com/v1/fonts.css">
+<script src="https://tikzjax.com/v1/tikzjax.js"></script>
+{% endif %}

_includes/scripts/misc.html

@@ -21,3 +21,6 @@
 <script src="{{ '/assets/js/clipboard.js' | relative_url }}"></script>
 <script src="{{ '/assets/js/copy_code.js' | relative_url }}"></script>
 
+<!-- Charts / diagrams / pseudocode (opt-in per page via front matter) -->
+{% include scripts/diagrams.html %}
+

_includes/search/ninja_keys.html

@@ -0,0 +1,44 @@
+<!-- Full-site command-palette search (ninja-keys). Opens with Cmd/Ctrl-K or the navbar icon. -->
+<ninja-keys placeholder="Search pages, posts, projects..." hideBreadcrumbs></ninja-keys>
+
+<script>
+  (function () {
+    var items = [
+      {% assign sorted_pages = site.pages | sort: "nav-title" %}
+      {% for p in sorted_pages %}{% if p.nav and p.autogen == nil %}
+      { title: {{ p.nav-title | default: p.title | jsonify }}, section: "Pages", url: {{ p.url | relative_url | jsonify }}, icon: "fa-file-lines" },
+      {% endif %}{% endfor %}
+      {% for post in site.posts %}
+      { title: {{ post.title | jsonify }}, section: "Blog", url: {{ post.url | relative_url | jsonify }}, icon: "fa-pen-nib" },
+      {% endfor %}
+      {% for pr in site.projects %}
+      { title: {{ pr.title | jsonify }}, section: "Projects", url: {{ pr.url | relative_url | jsonify }}, icon: "fa-diagram-project" },
+      {% endfor %}
+    ];
+
+    function mount() {
+      var ninja = document.querySelector("ninja-keys");
+      if (!ninja) return;
+      ninja.data = items.map(function (it) {
+        return {
+          id: it.section + ": " + it.title,
+          title: it.title,
+          section: it.section,
+          icon: '<i class="fas ' + it.icon + '"></i>',
+          handler: function () { window.location.href = it.url; },
+        };
+      });
+      var trigger = document.getElementById("search-toggle");
+      if (trigger) {
+        trigger.addEventListener("click", function (e) {
+          e.preventDefault();
+          ninja.open();
+        });
+      }
+    }
+
+    if (document.readyState !== "loading") mount();
+    else document.addEventListener("DOMContentLoaded", mount);
+  })();
+</script>
+<script type="module" src="https://cdn.jsdelivr.net/npm/[email protected]/+esm"></script>