Krish Patel

AI/ML Engineer specializing in Generative AI and agentic LLM systems for enterprise clients in shipping and tax. Delivers end to end, from client requirements and LangGraph agent design to LLM/VLM document extraction, Python APIs, PostgreSQL and deployment, with a focus on accurate, auditable AI output.

AI/ML Associate Software Engineer at Cygnet Infotech (Cygnet.One) (2026-01 – Present)

AI Laytime & Demurrage Automation | maritime shipping client Python, LangGraph, Azure OpenAI (GPT vision), PostgreSQL, Microsoft Graph API, pytest

• Delivered an end-to-end LLM agent system that automates laytime and demurrage calculation, replacing manual analyst work; owned it from client requirements to UAT and onboarded a second engineer.
• Designed an 8-stage LangGraph pipeline with LLM/VLM extraction agents for four shipping document types, routing digital PDFs through a layout-preserving text path and scanned ones through GPT vision.
• Built a deterministic calculation engine with a 40-rule validation layer and audit ledger; matched analyst results to the cent on all validation voyages and closed a six-figure gap to within 1%.

AI/ML Associate Software Engineer at Cygnet Infotech (Cygnet.One) (2026-01 – Present)

Litigation Management Solution | enterprise tax-litigation platform (GST, Income Tax) Python, Flask, PostgreSQL, SQLAlchemy, Alembic, Selenium, Microsoft Graph API, OAuth2, Swagger/OpenAPI

• Migrated email delivery for 6 enterprise client deployments from Outlook automation to Microsoft Graph API (OAuth2) as a drop-in replacement, with large-attachment uploads.
• Redesigned report-recipient storage, replacing three conflicting config sources with one PostgreSQL table and a rollback-safe Alembic migration, so clients can route GST and Direct Tax reports to separate teams.
• Engineered a Swagger-documented REST API for Direct Tax data deletion with PAN/TAN and run-level filters, dry-run mode and batched SQL for row deletes.
• Developing a resumable Selenium scraper for the Income Tax Compliance Portal that downloads only new notices and blocks any action that could modify taxpayer data.

AI/ML Associate Software Engineer at Cygnet Infotech (Cygnet.One) (2026-01 – Present)

GenAI R&D | internal proofs of concept Python, MCP (Model Context Protocol), Langflow, Pydantic

• Built tool-calling LLM agents on custom MCP servers, using Langflow for agent orchestration.
• Designed modular agent workflows that select the right MCP tool per request and return structured output.
• Implemented MCP tools with Pydantic schemas, input validation and logging for reliable backend calls.

Machine Learning Intern at CSIR – Central Road Research Institute (CRRI), Govt. of India (2025-02 – 2025-05)

• Built a computer vision pipeline in Python and OpenCV to detect and measure potholes and cracks.
• Integrated it into a road distress measurement tool and helped build its Flask interface for real-time results.

B.E. in Computer Engineering – LDRP Institute of Technology and Research, Gandhinagar (2022 – 2026)