Kirti K.

10+ years of enterprise software engineering, now applied to production AI. I’m a Senior AI engineer and Python developer building AI agents, RAG applications, LLM-powered automation, and AI chatbots that hold up under real workloads. That decade matters more than it sounds. I’ve spent it designing backend platforms, cloud-native applications, and distributed systems across healthcare, financial services, and enterprise software, working on systems that process critical business data in production environments where downtime and bad data have real costs. I bring that same discipline to AI. Most AI projects don’t fail because of the model. They fail on retrieval quality, hallucinations, missing guardrails, fragile integrations, and no way to tell when something breaks. Because I approach AI as a software engineer first, the system around the model gets the same care as the prompts: grounded answers with citations, validated structured outputs, human approval where it matters, logging, and evaluation before anything ships. Selected results: - Lead response automation (n8n, LangChain, HubSpot): response time cut from 19 hours to about 3 minutes, conversion up from 4.1% to 9.6%, and 87% of replies sent without human edits. - Invoice processing automation (FastAPI, Claude, AWS Lambda): 82% of invoices auto-posted at 98% accuracy, 34 duplicates caught, zero double payments. - Healthcare data quality platform: LLM and RAG anomaly detection across 38 data feeds and 12.4M records a day, with a HIPAA-aligned audit trail. - Production AI voice agent handling hundreds of thousands of calls, as part of the build team. More case studies are in my portfolio below. What I build: - AI agents and agentic workflows: multi-agent systems, tool and function calling, MCP, human-in-the-loop approval - RAG and LLM applications: OpenAI, Claude, Gemini, LangChain, LangGraph, embeddings, vector search (pgvector), prompt engineering, AI evaluation, guardrails - AI automation: n8n, Make, workflow automation, AI voice agents, HubSpot and CRM automation, lead enrichment with Clay and Apollo, email automation - Intelligent document processing: PDF and OCR extraction, validation, structured data - Backend and full-stack: Python, FastAPI, REST APIs, webhooks, Node.js, React, Next.js, TypeScript, PostgreSQL, Supabase - Cloud and data: AWS (Lambda, API Gateway, S3, Step Functions), Azure OpenAI, Docker, CI/CD, ETL pipelines, Databricks, PySpark, Snowflake - Machine learning: PyTorch, fine-tuning with LoRA, recommendation systems How I work: I’ve spent my career partnering with product owners, business stakeholders, and QA and engineering teams, and I work the same way with founders today: discovery, an honest recommendation (including when a simple deterministic automation beats AI), the build itself, and a documented handoff your team can run without me. A prompt is a request; a gate is a guarantee. Send me what you’re trying to automate and I’ll tell you honestly how I’d build it.