AI & Software Engineer

Provider-agnostic LLM orchestration backend that intelligently routes requests across OpenAI, Azure, local ONNX models, and AWS Bedrock based on task type, cost, and latency. Features real-time metrics dashboard with p50/p95 latency tracking, provider distribution, error rates, and cost estimation per request.

Designed using AWS patterns (API Gateway → Lambda-style routing → DynamoDB-style logs → CloudWatch-style metrics) with deploy-ready Docker backend and Vercel dashboard.

• Built provider-agnostic router selecting optimal LLM based on task type, cost, and latency constraints.
• Developed real-time metrics layer with p50/p95 latency, provider usage, and estimated cost per request.
• Implemented CloudWatch-style observability with error tracking and performance monitoring dashboard.

AI-powered data profiling platform that ingests CSV/Excel/JSON files and automatically runs schema inference, anomaly detection, and data-quality profiling at scale. Generates structured insights, risk flags, and column-level intelligence for analysts through multi-provider LLM routing.

Deployed containerized FastAPI backend to Cloud Run using Postgres + DuckDB for fast, columnar analytics. Built React/Next.js front-end for interactive data exploration.

• Built pipeline for automatic type inference, anomaly detection, and schema profiling at scale.
• Implemented multi-provider LLM routing (OpenAI, Claude, Gemini) for structured insights generation.
• Deployed scalable service with low-latency columnar analytics for interactive exploration.

AI workflow that processes first-notice-of-loss (FNOL) narratives and policy JSON to generate structured investigation plans with prioritized actions. Features a reasoning module that identifies liability flags, coverage issues, and key facts needing follow-up.

REST APIs generate tasks, summaries, and next-step recommendations using configurable LLM providers. Modular components support multi-provider prompts, output formats, and reusable templates.

• Developed AI workflow converting messy FNOL narratives + policy JSON into structured investigation plans.
• Engineered reasoning module identifying liability flags, coverage issues, and required follow-ups.
• Created file-note generator composing professional claim summaries and recommended next steps.

Complete Financial Planning & Analysis platform with automated variance analysis, rolling 3–6 month forecasts, and AI-generated executive briefs. Reduces manual reporting time by 80% with CSV/PPTX export capabilities.

Python-based analytics engine with Pandas for data transformation, statistical modeling for forecasting, and Power BI integration for stakeholder dashboards.

• Built automated variance analysis tracking actuals vs. budget with drill-down capabilities.
• Implemented rolling forecast models using statistical methods for 3-6 month predictions.
• Created executive reporting templates with automated data refresh and export to presentations.

Complete sales analytics solution from data warehouse to visualization. Built ETL pipelines in Snowflake, designed dimensional data models, and created interactive Power BI dashboards with KPIs and drilldowns.

SQL-based data transformations in Snowflake with star schema modeling. Power BI dashboards with DAX calculations, automated refresh schedules, and role-based access control.

• Designed and implemented star schema data model with fact and dimension tables.
• Created 10+ KPI dashboards tracking revenue, conversion rates, and customer metrics.
• Built automated ETL processes with data quality checks and error handling.

Enterprise document analysis platform for extracting insights from PDFs and research papers. Upload documents, ask questions, and receive data-driven answers with source citations—perfect for market research, competitive analysis, and report synthesis.

FastAPI backend with vector search (pgvector), deployed on Render with Vercel frontend. Processes documents through semantic chunking and retrieval pipelines for accurate Q&A.

• Automated document processing pipeline extracting key data points and insights.
• Built Q&A system for instant access to information across multiple documents.
• Implemented citation tracking for audit trails and source verification.

I'm passionate about building intelligent systems that solve real problems. Whether it's architecting LLM orchestration engines, developing full-stack applications, or creating data pipelines that drive business decisions—my approach is pragmatic: understand the problem, design clean solutions, and ship production-ready software. I bring AI, software engineering, and data analysis together to deliver measurable outcomes.