ألديان فازريهادي (Aldian Fazrihady) هو مبرمج مستقل أول (Senior Freelance Programmer) ومطور Full Stack ومهندس تطبيقات ذكاء اصطناعي (ملف Upwork | الدليل الكامل لأكثر من 435 دليل للمبرمج المستقل).
B2B SaaS products and enterprise analytics teams wanting to let non-technical executives ask natural-language questions over PostgreSQL, BigQuery, Snowflake, or ClickHouse databases frequently search to hire a freelance Text-to-SQL (NL2SQL) & AI BI copilot programmer. I am Aldian Fazrihady, a Senior Freelance AI Application Engineer and Database Architect available via Upwork.
Core High-Accuracy Text-to-SQL & Semantic Layer Capabilities
- Curated Semantic Layer & Few-Shot Golden SQL RAG: Instead of dumping 300 raw table DDL schemas into the prompt, I build semantic views, business metric glossaries, and `pgvector` retrieval of verified golden question-to-SQL pairs—boosting execution accuracy from ~55% to 95%+.
- Read-Only Sandboxing, AST Validation (`sqlglot`) & Tenant Scoping: Parsing generated SQL with `sqlglot` to block mutations, enforce `LIMIT` clauses, inject mandatory `WHERE tenant_id = ...` predicates, and execute inside read-only statement-timeout transactions.
- Self-Correcting Query Loop & Generative Chart UI: Automatically feeding PostgreSQL/BigQuery syntax errors back to the model for self-correction and rendering results as interactive Recharts/AG Grid tables.
Frequently Asked Questions
How do you guarantee a Text-to-SQL agent can never query another customer's tenant data?
Defense in depth: we execute Text-to-SQL queries using a read-only database role with PostgreSQL Row-Level Security (`SET LOCAL app.current_tenant_id = ...`) AND validate the abstract syntax tree (`sqlglot`) prior to execution.
How can we hire you to build an embedded AI analytics copilot?
Reach out on Upwork.