Yashwanth Y

Engineered production LLM-powered agent workflows with Python, FastAPI, and LangGraph, automating enterprise operations across 4 domains: employee information, task management, calendar events, and HR onboarding. Built workflow orchestration on Conductor OSS (task workers, workflow definitions, parallel execution); parallelising independent API calls cut observed latency from ∼3s to ∼1s (∼67% faster) in tested workflows. Developed LLM + RAG-based validation for meeting task extraction: identifies actionable tasks from transcripts, filters duplicate and already-assigned tasks, and rejects unsupported (hallucinated) suggestions (3 validation checks).

AI Engineer Intern at Aurumi.AI (2026-08 – Present)

• Engineered production LLM-powered agent workflows with Python, FastAPI, and LangGraph, automating enterprise operations across 4 domains: employee information, task management, calendar events, and HR onboarding.
• Built workflow orchestration on Conductor OSS (task workers, workflow definitions, parallel execution); parallelising independent API calls cut observed latency from ∼3s to ∼1s (∼67% faster) in tested workflows.
• Developed LLM + RAG-based validation for meeting task extraction: identifies actionable tasks from transcripts, filters duplicate and already-assigned tasks, and rejects unsupported (hallucinated) suggestions (3 validation checks).
• Implemented deterministic, Pydantic schema-validated email workflows that verify task assignments and assignee addresses before notifying, reducing unnecessary LLM calls.
• Designed scheduled due-date reminder workflows (3 triggers: before, on, and daily after the due date until completion) integrated with calendar APIs and Action Cards.
• Architected integrations for 2 accounting platforms (QuickBooks, Zoho Books) using Composio and PostgreSQL; explored canonical data models to normalise them.
• Tested read-only ETL workflows across 169 accounting/CRM records (11 entity types) with no production write-back.
• Debugged API integration and workflow-state failures; built reusable patterns for reliable agent automation.

AI Engineer Intern at Revive Battery BV (2026-03 – 2026-06)

• Built and deployed an LLM-based meeting task extraction system (Claude API, React, PostgreSQL) turning transcripts into structured tasks (owner, deadline, priority) stored in PostgreSQL.
• Designed IoT workflows (MQTT, ESP32) for real-time monitoring; built dashboards for multi-timezone teams.
• Added WebSocket real-time updates for live tracking, removing manual page refreshes.

Bachelor of Technology in Computer Science and Engineering – Malla Reddy University (2022-09 – 2026-06)

Intermediate (MPC) – Narayana Junior College (2020 – 2022)