Raghuraman (Raghu) • UT Austin MS, Class of 2026

NetApp • Machine Learning Intern

Applied ML | Forecasting | Regime segmentation | Product analytics

Working on website-visit signal analysis and product-usage trend prediction for adoption planning. The work centers on state-space forecasting, regime segmentation, and building decision-support outputs that product teams can actually use.

Sprinklr • Senior Forward Deployed Engineer

Conversational AI | Customer delivery | CI/CD | Production debugging | Demos

Led discovery, demos, implementation, and production debugging for customer-facing conversational AI deployments. Delivered end-to-end technical work across rollout, issue triage, and handoff, with a mix of product context and direct customer ownership.

Zifo RnD Solutions • Associate Consultant

Python | Genomics ETL | Data pipelines | Large-scale workflows

Built Python-based genomics ETL and large-scale data workflows for downstream analytics. The work was heavy on data pipeline reliability, workflow design, and getting structured datasets into usable shape for later analysis.

DraftForge Repair Agent

AkashML | FastAPI | Analyze-and-repair workflow | Patch generation | Static web studio

AkashML-themed code repair agent built around an analyze-and-repair workflow that inspects issues, generates patches, and exposes the flow through a FastAPI backend and static web studio. It is framed like a production-style developer tool rather than a one-off hackathon script.

• Implemented a code-analysis to patch-generation loop designed for iterative repair workflows.
• Structured the project around backend, UI, and CI-oriented roadmap thinking for production-style delivery.

LifeLedger

Data Portability Hackathon 2026 | Multi-source ingestion | Grounded insights | Explainable outputs

Personal finance intelligence engine built around multi-source ingestion, grounded insight generation, and explainable outputs. It is designed as a decision-support workflow that keeps recommendations traceable instead of relying on opaque summaries.

• Combined multiple portable data sources into a single finance-oriented intelligence workflow.
• Prioritized grounded and explainable outputs so insights remain inspectable for end users.

ChargePilot EV Optimizer

Python | ETL pipelines | Graph message passing | Facility-location optimization | Next.js | GeoJSON

Geospatial AI decision-support system that ingests real mobility and infrastructure datasets, learns site scores, and ranks EV charger expansion scenarios through a map-based interface. It combines ETL, learned scoring, graph message passing, and facility-location optimization into a deployment-ready recommendation engine.

• Implemented an end-to-end pipeline from public urban data through scored recommendations and scenario artifacts.
• Produced map-ready outputs for interactive decision support instead of stopping at a notebook-only analysis.

Smart Doc Approver

LangGraph | OCR | LayoutLMv3 | Anomaly detection | Approval routing | Human-in-the-loop

Agentic receipt automation system that combines OCR, document extraction, anomaly detection, and approval routing for production-style document workflows. It routes uncertain cases through human review and uses feedback loops to improve extraction quality over time.

• Unified extraction, validation, and confidence-based routing so low-confidence cases do not silently fail.
• Closed the loop with review feedback to drive repeated model-quality and validation-quality improvements.