I turn messy, fast-moving data into clear decisions. My sweet spot is end-to-end work: scoping the question, cleaning the data, finding the signal in Python/SQL, and shipping something real—dashboards people use or a lightweight ML/LLM feature that actually moves a metric.Recent wins: built 7 executive KPI dashboards used by 12 stakeholders and shaved ~3 hours off weekly reporting; shipped a sales-commitment scoring system (LLM + ML on ADK Web) that flags over/under-committing with ~1.2s median latency and 89% CV accuracy (~220 pilot runs); designed a box-count predictor (Optuna-tuned RF, RMSE 0.30 on ~50k rows) and put it in users’ hands via Streamlit + ADK; prototyped an agent-to-agent Weather Q&A that answers in under a minute so planning teams don’t hunt across sites.What this really means is I can own the path from vague problem → measurable result: SQL and pandas/NumPy for analysis, scikit-learn/XGBoost for modeling, Streamlit/ADK for quick deploys, and Looker Studio/Tableau/Power BI to tell the story without the fluff. I care about clarity, reproducibility, and honest metrics—latency, accuracy, adoption.Toolbox: Python (pandas, NumPy, scikit-learn, XGBoost), SQL/BigQuery, Looker Studio, Tableau, Power BI, Streamlit, LLMs (prompting + eval), Google ADK Web, Git, basic Docker.Open to Data Analyst / Junior ML / Analytics roles. If you need someone who can ship a dashboard this week and an ML/LLM MVP next week, let’s talk.
This is the second paragraph of your amazing article.
This is the third paragraph where the content continues.
• Developed 7 executive KPI dashboards in Looker Studio adopted by 12 stakeholders; reduced weekly
reporting prep by ~3 hours via self-serve access.
• Delivered 4 production-ready classifiers/regressors across tabular, text, and image data (best models ~90%
accuracy) with clear eval and handover docs.
• Designed a box-count predictor (Optuna-tuned Random Forest, RMSE = 0.30 on ~50k rows) that
standardized decisions and cut packing mistakes by ~15%; deployed via ADK Web and Streamlit for direct
usage.
• Delivered a sales-commitment feasibility system (LLM + ML) with Google ADK Web: CV accuracy 89%;
prompt-driven scoring classifies over-committing / under-committing / good-to-go; median response ~1.2
second; ~220 pilot runs. Stack: Python, scikit-learn, ADK Web, Streamlit.• Built an AI Weather Bot (agent-to-agent) with ADK Web for real-time Q&A; less than 2 second end-to-end
responses, removing manual checks during planning.
• Built 3 ML prototypes for ticket categorization and routing; reduced manual triage time by ~20% in pilot
evaluations.
• Produced 14 weekly trend reports on incident/request volumes; surfaced 3 recurring SLA-breach patterns
for ops leads.
• Ran SonarQube/OWASP ZAP reviews; analyzed scan data to surface top vulnerability categories and severity
trends; built a Power BI dashboard to present findings to stakeholders; documented 12 remediation actions
adopted by engineering for safer releases.
• Recommended 5 data-backed process tweaks (queues, hours, SLAs) that cut backlog by ~12% during trial.
• Built competitor analysis using SQL/Excel (Power Query, pivots) across five competitors with pricing and
channel coverage.
• Ran marketing time-series tests in R on six months of data to guide budget allocation and campaign timing.
• Created Tableau views to track spend vs. ROI; shared weekly insights with leadership for decision-making.
• Applied basic ML in R on ~5,000 customer rows to support demand forecasting and inventory planning.
• Built a Power BI dashboard for four core metrics; assisted SQL/Excel migration of ~10,000 records with data-
quality checks.
• Co-authored three summary reports and dashboards for management decision-making.
Restricted contact info. You need company account to have access.
