Jaswanth Sirigiri
I build reliable, cloud-native data pipelines that turn raw, messy data into trusted, AI-ready assets. 5+ years across financial services and healthcare, with deep focus on Azure, Databricks, and Snowflake.
That same instinct for clean, well-governed data carries into my project work: fine-tuning models, building retrieval and explainability layers, and shipping ML-powered tools on top of the pipelines that feed them.
- Cloud Platforms
- Azure (ADF, Synapse, Purview, DevOps, Monitor, Logic Apps, ADLS Gen2, Blob), AWS (S3, Glue, Redshift, Lambda)
- Big Data & Streaming
- Apache Spark, PySpark, Databricks, Delta Lake, medallion architecture, Apache Kafka
- ETL / ELT
- Azure Data Factory, AWS Glue, CDC, incremental loading, event-driven pipelines
- Warehousing & Modeling
- Snowflake, Synapse Analytics, dimensional modeling, star schema, partitioning, clustering
- Databases & Formats
- SQL Server, PostgreSQL, MongoDB, CosmosDB, Parquet, Avro, JSON
- Quality & Governance
- Validation frameworks, lineage, Azure Purview, anomaly detection, HIPAA, masking, audit logging, SLA/SLO
- Languages
- Python, SQL, PySpark, Shell Scripting
- DevOps
- Git, Azure DevOps, Docker, CI/CD automation
- Reporting & AI
- Power BI, Tableau, Anthropic Claude API, LLM-assisted root-cause analysis
A few things I've built along the way, solo and with teams.
A medallion-lakehouse pipeline for healthcare claims with automated data quality checks and AI-generated root-cause triage.
A graduate team project designing an AI finance assistant with stock forecasting, sentiment analysis, and a conversational agent.
A Streamlit app analyzing 7,043 customers with churn KPIs, EDA, a prediction model, and GenAI-generated retention insights.
A graduate HCI project building an emotion-aware chatbot: a fine-tuned RoBERTa model paired with a purpose-built chat interface.
A multimodal AI system that detects whether an image is AI-generated or human-captured, combining ViT classification, FAISS retrieval, and T5-based natural language explanations.
A team research project applying machine learning to detect DDoS attacks in Software-Defined Networks, training and comparing five models on a 104K-record traffic dataset across UDP, TCP, and ICMP protocols.
Two years going deeper into the systems side of computer science: distributed architectures, human-computer interaction, and machine learning, building on years of hands-on data engineering work with the kind of formal grounding that makes the next hard problem a little less unfamiliar.