I work with data end to end - pulling and modeling it is SQL, analyzing it in Python, and turning it into dashboards in Power BI. Recently I've been applying that same approach to RAG and LLM systems: not just calling an API, but measuring retrieval quality, tracking latency and token cost, and building the pipelines that make an LLM application observable and reliable.

RAG application over a 1,979-document corpus, with retrieval strategies benchmarked head-to-head against 1,881 ground-truth questions.

• Compared Text Search, TF-IDF, and Semantic Search (Sentence Transformer embeddings) — Text Search led at 99.95% Hit@5 / 99.73% MRR
• Modular pipeline: ingestion → retrieval → prompt construction → LLM generation → evaluation → monitoring
• Llama 3.3 70B via Groq, with a Streamlit dashboard tracking token usage, latency, and user feedback
• Fully containerized with Docker for reproducible deployment

Stack: Python · RAG · Groq · Sentence Transformers · dlt · DuckDB · Docker · Streamlit [Link]

Instrumented a RAG pipeline with distributed tracing to answer a question most RAG demos never ask: where does the time and cost actually go?

• OpenTelemetry spans across rag, search, and llm stages — found LLM inference dominates latency (~12.9s vs ~4.5ms for search)
• Retrieval evaluated with Hit Rate/MRR; answer quality scored via cosine similarity and LLM-as-a-Judge
• Built a SQLite span exporter and a dlt → DuckDB pipeline to query trace data with SQL
• Added Pydantic AI + Logfire for agent-level observability (per-execution span tracking)

Stack: Python · OpenAI/GitHub Models · PostgreSQL · Grafana · OpenTelemetry · Logfire · dlt · DuckDB [Link]

More projects in my pinned repositories.

• Languages & Querying: Python · SQL
• Data Engineering: dlt · DuckDB · PostgreSQL · SQLite · Docker
• BI & Visualization: Power BI · Microsoft Fabric · Streamlit · Grafana
• AI / LLM: RAG · Vector Search · LLM Evaluation · OpenTelemetry · Logfire · Pydantic AI
• Tools: Git · GitHub · Jupyter · NumPy · Pandas

• Microsoft Certified: Fabric Analytics Engineer Associate
• Microsoft Certified: Power BI Data Analyst Associate

• Master of Computer Applications — ITS Mohan Nagar, Ghaziabad Affiliated to Dr. A.P.J. Abdul Kalam Technical University, Lucknow | 2025 - 2027
• Bachelor of Computer Applications — DIT University, Dehradun | 2021 - 2024