Forward Deployed AI Engineer & Tech Lead, HUMAI Institute
02/2025 - Present · Contractor
- Single point of contact for clients: requirements gathering, weekly meetings, demos, handoff, and interviewing engineers for staff augmentation
- Coordinated teams of external AI experts to produce cutting-edge AI courses: interviewed and negotiated with instructors, wrote scripts and teaching guides, and ran QA on every recorded class
- Built full-stack applications with FastAPI, PostgreSQL (Cloud SQL), GCP (Cloud Run, GCS), GitHub Actions, and Docker
- Implemented secure RBAC, cookie-based authentication, and real-time scraping with Playwright
- Designed ETL pipelines with VertexAI (LLM-based and RAG) for data processing and analysis
- Developed React frontend with SSE, Kotlin companion app, and backend using FastAPI with CI/CD
- Taught 100+ classes on Python, automation, web scraping, databases, CI/CD, and Git
Client projects
Conversational AI Interview Platform
LLM-powered conversational interview platform for qualitative research: an AI agent conducts in-depth interviews by text or voice, decides when to probe deeper, and consolidates findings. Ran discovery, biweekly demos, and handoff directly with the research team, and selected models through LLM-as-judge evals on real interviews. Built the complete system end-to-end on AWS.
AI, Next.js, TypeScript, Claude, AWS Bedrock, PostgreSQL, AWS
Parental Monitoring Platform
LLM-powered parental monitoring application that scrapes messaging platforms using whatsmeow and processes conversations through multiple LLM layers to generate emotional state metrics and alerts. Worked directly with the founders from beta to funding. Built the complete system end-to-end: backend, frontend, scraper, LLM analysis service, alerting, and observability with Langfuse on GCP.
AI, Go, whatsmeow, OpenAI, Claude, Langfuse, GCP, React, FastAPI
Music Industry Management App
AI-driven application for music managers that uses LLMs to generate and execute artist launch plans. Worked directly with the founder, led a team of 3 engineers, and owned model and architecture decisions: migrated Django to FastAPI with zero downtime and cut LLM cost per plan by 88%.
AI, React, FastAPI, Django, OpenAI, PostgreSQL
Government Process Automation
Automated remote processing system for government registry procedures, replacing manual filing by the client's operations team. Gathered requirements with operators and the PM, directed the client's technician, and put into production a self-healing agent pipeline that recovers lost sessions, resumes from the failed step, and never files twice. Built with Playwright, backend API, dashboard, and batch processing.
AI, Playwright, FastAPI, React, PostgreSQL
Banking Data Migration
Senior Data Engineer migrating a daily reconciliation process from on-prem IBM AS400 to the cloud using AWS, Databricks, Airflow, Spark, Kafka, and Kubernetes. Led the 10-person team's AI adoption: Claude Code skills, hooks as guardrails, and secure MCP servers to on-prem and observability tools. Optimized ETL pipelines and DAGs, reducing processing times by over 50%.
AI, AWS, Databricks, Airflow, Spark, Kafka, Kubernetes, Python