Priyanshu Shekhar Sinha

Lead the architecture and development of Discover, an enterprise RAG platform spanning ingestion, retrieval, ranking, agentic reasoning, and access control across cloud providers.

• Designed and shipped a ReAct-style reasoning agent for natural-language analytics — surfaces key insights as the final answer, generates large-tabular insight payloads, and harvests delegated-agent and web references into inline citations; added an orchestrator / sub-agent role toggle to the agent canvas.
• Built a stateless Streamable-HTTP MCP server exposing retrieval as tools (ask, search, aggregate_jira/confluence), secured with OAuth 2.1 resource-server auth (Keycloak JWT) and Kubernetes liveness probes.
• Engineered a staged connector-ingestion pipeline (fetch → Docling parse → contextualise → embed) with windowed processing and per-knowledge-base tuning presets; delivered Jira, Confluence, S3 and SVN connectors behind a unified experience.
• Researched and implemented Citus DB for distributed, enterprise-scale Postgres workloads and migrated the backend to a production gunicorn multi-worker server.
• Onboarded the first enterprise client for the Text2SQL solution; mentor a cross-functional team on data pipelines, LLM fine-tuning, and deployment strategy.

Foundational AI engineering on Discover — Elastiq's flagship product — covering fine-tuning, retrieval, and agent design.

• Fine-tuned LLaMA 3.1 70B via advanced QLoRA and LoRA techniques, reaching 85% target accuracy.
• Built the foundational Text2SQL framework from scratch and shipped AI agents for natural-language-to-SQL plus unstructured file processing.
• Integrated multiple Graph DB backends offering Classic RAG, Cypher-based RAG, and GraphRAG — extending Discover with structural retrieval.
• Engineered custom datasets and context-aware chunking algorithms to lift unstructured-document RAG performance.

Research, analytics, and technology partner to the financial-services sector — asset managers, investment banks, private equity, and hedge funds.

• Developed a versatile analysis dashboard for an investment firm enabling KPI monitoring and experimentation — 87% reduction in client analysis time.
• Built a web dashboard integrated with an ESG data engine for a top-tier investment firm, shaping their investment-strategy decisions.
• Shipped an end-to-end RAG system for parsing financial reports.

• Led a 5-person team shipping a flexible A/B testing framework for a leading pharmaceutical client, measuring promotion success against revenue lift.
• Co-led a 7-person team on a seasonality-aware sales-anomaly pipeline that drove customer outreach delivering +$125K revenue in Q1.

• Cut new-client integration from 10 days to 3 hours (97%); operational reports from 5 days to 30 minutes; opportunity reports from 1 day to 30 minutes.
• Reduced advanced-analytics report generation from 7 days to 15 minutes (99%) and automated Power BI dashboard refreshes.

• Designed a multi-phase pipeline enriching extracted document data points — generated 500+ features via NLP, data wrangling, and text mining.
• Co-developed a matrix-multiplication ensemble on the NumPy stack across 10+ sources for entity classification, plus custom Bayesian models for word embeddings.