Skills

Core engineering stack, AI frameworks and infrastructure tools.

LLM & AI Agents

pydantic-ai & Pydantic Graph

Spring AI

LLM API Integration (OpenAI, Gemini, Anthropic)

Function / Tool Calling

Structured Output

Prompt Engineering

HyDE & Query Decomposition

LLM-as-judge

RAG & Retrieval

RAG Architecture

Document Ingestion (PDF, Word, Excel)

Chunking Strategy

Embedding Models

Vector Search (Qdrant, pgvector)

Hybrid Search (BM25 + Vector, RRF)

Reranking (LLM, Cross-Encoder)

Metadata Filtering

Document AI & Evaluation

OCR (PaddleOCR, Tesseract, Google Vision)

Vision LLM Extraction

PP-OCR Fine-tuning

OCR Benchmark (Exact Match, CER/WER)

LLM Output Evaluation

AI Safety & LLM Ops

Prompt Injection Mitigation

Secret / Sensitive Data Redaction

Document-level Access Control

Multi-LLM Failover & Rate Limiting

Tracing & Logging

Token Budget & Cost Tracking

Backend

Python

FastAPI

Celery

Java

Spring Boot

Spring Cloud Gateway

Spring Security

Spring Data JPA

Databases

PostgreSQL

pgvector

Qdrant

MySQL

MongoDB

Redis

Cloud / DevOps

AWS

Docker

Kubernetes

CI/CD (Jenkins)

Kafka & MessageMQ

MinIO

Git

Frontend

JavaScript

TypeScript

React

Next.js

React Native

Architecture

Microservices

Monolithic Architecture

REST APIs

WebSocket

Server-Sent Events (SSE)

JWT Authentication

Active Focus

Continuously iterating on retrieval mechanisms: hybrid search (BM25 + vector with RRF) and cross-encoder reranking for UniSage.