I build AI applications backed by reliable data systems. My foundation is approximately four years in data engineering at Everlytics, working on enterprise migrations, ETL pipelines, SQL optimization, and production support. I now apply that experience to independent projects in document retrieval, tool-using agents, and natural-language data access.
Open to Applied AI, Python + GenAI, and AI Data Engineering roles in Delhi NCR.
Portfolio & case studies · LinkedIn · Email · Repositories
These are independent portfolio projects. My commercial production experience is in data engineering; project links below show the implementation, tests, and current scope of my AI work.
| Project | What it does | Engineering evidence |
|---|---|---|
| Document QA & RAG | Answers questions over PDFs with file/page citations using LangChain, FAISS, Ollama embeddings, and Groq. | Index-integrity manifests, modular query services, Docker Compose, and automated tests. The offline evaluation fixture checks regressions, not real-world answer quality. Quality gates |
| AutoDocsGenAI | Develops a documentation ingestion and retrieval pipeline from official sources using Python, PostgreSQL, and pgvector. | Deterministic chunking, persisted lineage, repeatable imports, and embedding/search/evaluation commands. The project is in progress; generation and review remain in development. Architecture |
| Agentic AI Demo | Demonstrates bounded tool use, supervisor/specialist workflows, LangGraph checkpoints, and MCP in a support-operations scenario. | Explicit approval policy, validation, loop limits, traces, and offline tests. Uses dummy data and in-memory state. Code and tests |
| AI SQL Assistant | Provides a natural-language interface to SQLite, MySQL, and PostgreSQL through LangChain, Groq, and SQLAlchemy. | Schema discovery, input checks, read-only SQLite access, and response-time tracking. Application-level SQL checks are prototype safeguards, not a security boundary. Screenshots and design |
For a quick review: start with Document QA for applied RAG, AutoDocsGenAI for data architecture, or Agentic AI Demo for orchestration and tool policies.
At Everlytics, I progressed from Data Engineering Intern to Data Engineer and Senior Data Engineer, working on warehouse migrations, ETL delivery, and production support.
- Contributed to a Teradata-to-SingleStore migration spanning 100+ TB, and migrated or optimized 100+ Informatica workflows.
- Coordinated Airflow workflows across batch, CDC, and streaming ingestion, with validation, dependencies, monitoring, and release handover.
- Owned a file-to-SingleStore pipeline from development and testing through production: compressed files arriving every five minutes, shell-based validation, landing/staging loads, SCD Type 2 dimensions, fact loading, rejected-record handling, task-level alerts, and archival.
- Investigated failed loads and data discrepancies, tuned SQL and workflows, and communicated delivery risks and recovery actions to stakeholders.
That experience shapes how I build AI systems: inspect failures, preserve data lineage, make workflows testable, and document operational limits.
| Area | Tools and practices |
|---|---|
| Python & APIs | Python, FastAPI, REST APIs, SQLAlchemy, modular design, input validation |
| Applied AI | RAG, embeddings, FAISS, pgvector, prompt/context design, tool calling, LangChain, LangGraph, Groq, Ollama |
| Data systems | Python, SQL, Airflow, Informatica, PostgreSQL, SingleStore, Teradata, ETL, CDC, dimensional loading |
| Quality and delivery | Pytest, GitHub Actions, Docker, type checking, retrieval evaluation, logging, LangSmith, root-cause analysis |
- Better retrieval and answer-quality evaluation on representative documents.
- Clear API boundaries and reproducible deployment for portfolio AI applications.
- Reliable agent execution with explicit permissions, bounded state, and measurable failure handling.
- Sharing project architecture and engineering lessons on my portfolio.
These badges reflect repository activity and workflow status. The project documentation explains what each check covers.
Interested in engineers who can connect enterprise data experience with applied AI? I would welcome a conversation about your team's work.
Connect on LinkedIn or email me for my latest resume and project walkthroughs.