Building Data & Agentic AI Systems
Data pipelines, RAG systems, and LLM workflows — end to end, deployed, and documented.
M.S. in IT & Management, UT Dallas '26 — Dean's Excellence Scholar · ex-HCLTech · Dallas, TX
prasannawarad.com · LinkedIn · Email
🟢 Open to full-time roles — Data Engineer · AI/ML Engineer · Data Platform
Based in Dallas, TX · open to relocation anywhere in the U.S. · available immediately
🤖 Don't feel like scrolling? Ask Prasanna AI — the chat widget on my portfolio answers questions about my work, my projects, and even my anime backlog. I built it myself: Cloudflare Worker + Groq, grounded in a structured knowledge base.
| If you're hiring for… | Read these first |
|---|---|
| Data Engineering / Platform | market-research-copilot — Spark → Delta/Unity Catalog → pgvector serving · CardioRisk — deterministic config-driven pipeline · plus production Spark/Airflow reporting pipelines for national retail operations at HCLTech ↓ |
| AI / ML Engineering | sec-rag-intel — RAG with RAGAS-measured faithfulness · RAGbase — hybrid BM25 + vector retrieval · credit-risk-default-prediction — imbalanced-class modeling |
| Full-stack / Product | codelens-ai · InvestIQ · DataDoc_AI — all three are live and clickable |
| Project | What it does | Stack |
|---|---|---|
| sec-rag-intel | Production-grade RAG over SEC 10-K/10-Q filings — grounded, cited answers built on MMR retrieval and local BGE embeddings, with hallucinations caught quantitatively via RAGAS faithfulness scores | Python · LangChain LCEL · ChromaDB / Pinecone · Groq Llama 3.3 70B · FastAPI · Streamlit |
| codelens-ai — live demo | AI code-audit & technical-debt tracker: deterministic static metrics × LLM analysis, async BullMQ job queue, incremental re-audits via content-hash diffing, weighted 0–100 debt score over time | Node/Express · React · BullMQ + Redis · Prisma + Postgres · Gemini |
| RAGbase — live | Document-intelligence RAG platform: hybrid retrieval (BM25 + vector + RRF), streaming answers with clickable chunk-level citations, chunk inspector, analytics dashboard | Next.js · Supabase pgvector · Gemini embeddings · Groq (SSE) |
| InvestIQ — Top 5, Goldman Sachs × UTD JSOM Hackathon — live | AI investing guide for beginners: deterministic portfolio engine, streaming AI copilot with real-time voice briefings, and a Chrome extension that answers "what does this news mean for my portfolio?" | Next.js monorepo · Supabase · Groq · ElevenLabs · Plasmo MV3 |
| DataDoc_AI — live | Data-quality debugger: profile a CSV entirely in-browser, get AI-explained issues with suggested fix SQL, and query it with natural language → SQL | React · in-browser SQL engine · Express proxy · Gemini |
| prepai-pro — live | Interview intelligence: generates a company research dossier + STAR stories from your resume, then runs a scored multi-turn mock interview with adaptive difficulty and Whisper voice answers | React · Vite · Gemini 2.5 Flash · Groq Whisper |
| Project | What it does | Stack |
|---|---|---|
| market-research-copilot | Lakehouse-to-agent pipeline: Spark job computes rolling returns, volatility, volume z-scores, and drawdown via window functions, embeds article text with a pandas UDF, and writes partitioned Delta + a distributed upsert into Postgres with pgvector/HNSW — then an MCP server exposes 10 agent tools, 4 of which write back | PySpark · Delta Lake / Unity Catalog · Lakebase Postgres · pgvector · FastMCP · Flask |
| credit-risk-default-prediction | Default prediction on 255K+ loan records with ~12% positives: SMOTE inside stratified 5-fold CV, RandomizedSearchCV model comparison, RF threshold calibration, partial-dependence interpretation | Python · scikit-learn · XGBoost · imbalanced-learn |
| CardioRisk | Config-driven, fully deterministic ETL + ML pipeline over ~300K healthcare records — modular stages, reproducible seeds, headless CLI execution; champion model at ~91% accuracy / 0.81 AUC | R · tidyverse · caret · randomForest |
| Scalable Fleet Risk Analytics | Hadoop-based ingestion and transformation for multi-state fleet operations with risk and compliance monitoring | Hadoop · HDFS · MapReduce · Power BI |
| Instacart Customer Behavior | Processed 3M+ grocery transactions into analytics-ready datasets with KPI dashboards for demand and ROI tracking | SQL · Python · Tableau |
Also: My_Portfolio — the source behind prasannawarad.com, including Prasanna AI, served by a Cloudflare Worker + Groq.
Spark · Airflow · Databricks · Delta Lake · Unity Catalog · Kafka · Hadoop · dbt · Snowflake · Redshift · Azure Synapse · Oracle
scikit-learn · MLflow · LangChain · LangGraph · RAG / pgvector / ChromaDB / Pinecone · RAGAS · LLM Fine-tuning · Multi-Agent Systems · MCP / FastMCP
Groq · Gemini · Tableau · Power BI · REST APIs · Cloudflare Workers · n8n · BullMQ · Prisma
simple > clever ──── every architecture decision should survive an interview whiteboard
pipelines, not scripts ── window functions and partitioned writes, not a for-loop in a Spark wrapper
free tier, prod patterns ── quotas are real constraints → every LLM call has a fallback chain
grounded or it didn't happen ── RAG answers ship with citations + RAGAS faithfulness scores
deployed > demo ──── live URL, seeded demo account, screenshots — or it's not done
deterministic core ── LLMs decorate the edges; the engine underneath is reproducible math
| Role | Focus | Time |
|---|---|---|
| Software Engineer Intern, Cloud BC Labs | Built agentic AI hiring automation — REST APIs with Node.js and Python, ETL optimization, behavioral trend analysis | Oct 2025 – May 2026 |
| Data Engineer, HCLTech (Client: Dollar General) | Spark + Airflow reporting pipelines for a national retail network, large-scale transaction analysis, anomaly forecasting on AWS logs | Feb 2023 – Jul 2024 |
5-stage hiring process automated end-to-end with LLM agents
Top 5 Goldman Sachs × UTD JSOM Hackathon — InvestIQ
2M+ retail transaction records analyzed to improve reporting data quality
28% query latency reduction in real-time reporting pipelines
300+ fleet assets monitored across 14 states in risk analytics workflows
3M+ grocery transactions processed for segmentation and demand insights
- The University of Texas at Dallas — M.S. in Information Technology & Management (May 2026)
Dean's Excellence Scholar · GPA 3.88 · Graduate Certificate in Business Analytics & Data Mining - Savitribai Phule Pune University — B.E. in Electrical Engineering
┌─ weekends ────── cricket · pickleball · table tennis · soccer
├─ weekly ──────── soccer matches (haven't missed a week yet)
├─ nightly ─────── anime · Marvel · Suits · Billions · Modern Family reruns
├─ ongoing ─────── fiction & philosophy reading
└─ on trigger ──── random idea → side project → somehow becomes a real tool
Since completing my M.S. and my internship at Cloud BC Labs in May 2026, I've been deepening my lakehouse and agentic-AI engineering through structured boot camps and shipping the results publicly.
- Databricks lakehouse boot camp — Spark on Delta Lake / Unity Catalog, Lakebase Postgres, pgvector serving, and MCP tool servers; capstone shipped as market-research-copilot
- Production RAG & LLM evaluation — grounded answers, cited sources, faithfulness measured rather than assumed
- Shipping end to end — six of the projects above are deployed and publicly accessible, not screenshots in a slide deck
Open to full-time Data Engineer / AI/ML Engineer / Data Platform roles. Based in Dallas, TX and willing to relocate anywhere in the U.S. Reach me at waradprasanna@gmail.com or on LinkedIn.