| Languages |
|
| AI, RAG & NLP |
|
| Data & MLOps |
|
| Backend & Web |
|
- π§ Production RAG & LLM Orchestration: Architecting resilient semantic search and RAG systems with dense vector indexing, hybrid ranking, and prompt optimization (~70% low-latency retrieval boost).
- β‘ Data Pipelines: Streaming ingestion (Kafka + PySpark Structured Streaming with watermarking), SCD Type 2 modeling, and data-quality checks with quarantine and observability (portfolio projects with a built-in data simulator).
- π οΈ Backend Services: TypeScript/Node.js (Express) and Go services on PostgreSQL with versioned migrations, Docker and end-to-end tests; open-source Go contributions (8 merged pull requests in go-magnetar).
- π Multilingual NLP & LLMs: Cross-lingual language processing, tokenization, semantic similarity, and domain-adapted transformer fine-tuning.
| Repository | Focus & Architecture | Domain |
|---|---|---|
π my-rag-chatbot-3 |
RAG chatbot on LlamaIndex: Chroma vector store, BM25 + vector hybrid retrieval with reranking, Ollama / OpenAI-compatible LLMs, and a QA-generation evaluation module (personal project) | AI / LLM |
β‘ Data-Quality-and-Observability |
Automated data quality assertions, pipeline telemetry & metrics validation (runs on a built-in data simulator) | Data Engineering |
π Real-time-Clickstream-Anomaly-Detection |
Kafka + PySpark Structured Streaming anomaly detection on a built-in synthetic-event simulator (5-min window, 10-min watermark) | MLOps / Streaming |



