AI Engineer (Founding) at EditLingo Solutions LLC (2025-08 – 2026-04)
HistOracle (historacle.ai) — multi-tenant B2B SaaS platform for conversational learning with historical figures, taken from zero to production as one of three founding engineers.
- Architected a production multi-tenant RAG system on GCP with dual RAG pipelines combining dense semantic retrieval and hybrid BM25 reranking, ingesting 145K+ document chunks across 6K+ source documents into Pinecone with per-org namespace isolation, serving live users at Henry George School of Social Science
- Engineered an 8-phase batch document processing pipeline (GCS, extract, chunk, embed, upsert) with real-time SSE progress tracking, reducing document ingestion from a manual process to a fully streamlined admin workflow
- Reduced OpenAI API costs by 73% through prompt caching with an 80-85% cache hit rate, cutting cost/request from $0.015 to $0.004 and first-token latency from 6.5s to under 2s
- Developed a multi-agent Journey Mode system with a 6-phase stateless orchestrator and LLM-based checkpoint evaluation, delivering 27 published immersive learning experiences across 18 historical characters
- Devised background audio pre-generation with an LRU lock pool to prevent duplicate TTS calls under concurrent access, cutting perceived per-checkpoint latency by ~2s in Journey Mode
- Designed a LangGraph-orchestrated multi-character debate engine supporting 5-round real-time debates, using SSE token streaming to reduce time-to-first-token from 60+s to under 2s
- Built an asyncio-based user participation checkpoint system for the debate engine with zero CPU overhead during wait states, alongside a Learning Hub with continuous intent detection across 6 adaptive learning modes
- Shipped full CI/CD on GCP Cloud Run with GitHub Actions covering 3-stage sequential deployment (migrations, backend, frontend) with canary releases, 8-retry health checks, and automatic rollback across 103+ database migrations
Senior Data Analyst at QX Global Group (2022-01 – 2023-07)
- Launched a spend data analytics service from scratch, applying ML models (Regression, Naive Bayes) and delivering executive BI dashboards in Tableau and Power BI, reducing data processing turnaround time by 60%
Senior Data Analyst at Zycus Infotech Pvt. Ltd. (2020-10 – 2021-10)
- Led a team of 5 analysts delivering $250K+ in annual savings for Fortune 500 clients through spend analytics, earning a perfect 10/10 client satisfaction score
Data Analyst at Zycus Infotech Pvt. Ltd. (2018-07 – 2020-09)
- Engineered end-to-end spend data analysis for Fortune 500 clients using SQL and K-Means clustering on NLP-derived text features, automating procurement categorization and reducing processing cycle time by 40%
