// Skills & Proficiency
Core competencies across the data science and ML stack.
Technical Proficiency
Capability Radar
// Featured Projects
Dashboards, models, and pipelines built for real use cases.
Streamlit Analytics Dashboard
Multi-tab analytics app with DuckDB in-process querying, Plotly interactive charts, CSV/Excel export, and real-time data refresh.
Local LLM Fine-Tuning Pipeline
QLoRA fine-tuning on RTX 4080 using HuggingFace PEFT + TRL. Supports Qwen2.5-Coder-14B with 16GB VRAM constraint. Docker-portable adapter.
MS Teams Analytics Service
Node.js + DuckDB service processing Teams export data. Parameterized queries, RCE-safe API design, automated chart generation via workers.
GitHub Activity Visualizer
D3.js + Chart.js visualizations of GitHub repo stats — contribution heatmaps, language distributions, and growth curves from live API data.
AI Cost Optimization Orchestrator
Multi-provider AI routing system with cost waterfall logic. Classifies tasks and routes to optimal provider — local, free cloud, or paid API.
PyTorch CUDA Benchmark Suite
RTX 4080 benchmark suite for model inference speed. Tests batch sizes, precision modes (FP16/BF16), and memory throughput. Outputs JSON reports.
Message Volume by Channel
NLP Category Distribution (15 Domains)
Weekly Activity Trend
User Engagement Heatmap — Hour × Day of Week
Low
High
Snowflake Query Performance (ms)
Jira Sprint Burndown
Confluence Page Views Heatmap
Multi-Model Forecast Comparison
Model Accuracy (MAE)
Training Time (s)
QLoRA Fine-Tuning — Loss & Validation Curves
VRAM Usage by Batch Size
Tokens / Second by Precision
Column Data Type Distribution
Null / Missing Values by Column (%)
Cardinality & Uniqueness Profile
// Tech Stack
Tools and frameworks across the full data lifecycle.
Python 3.14
PyTorch
Plotly
DuckDB
Streamlit
HuggingFace
Chart.js
D3.js
Pandas
NumPy
scikit-learn
Docker