AI / Software Engineer · Ex-Atlan
what I build
50 customers · 14 days
up from 50
~450× faster
Work Experience
All four ran for paying customers at Atlan, where I was Software Engineer II, AI from Feb 2024 to Jun 2026. The numbers are measured, not projected.
200+
tickets/mo
deflected, up from 50
Support-automation system on Zendesk: doc-grounded triage over Glean, specialist sub-agent routing, and human-escalation workflows.
2×
first-pass correctness
≈40% → ≈75% on benchmark QA
In-house platform where an autonomous code-generation agent takes an engineering ticket and ships the functionality end-to-end.
70%
less effort
to document a metric
Auto-discovers KPIs from BI tools and SQL lineage, converts them into governed business metrics, and publishes a tenant-wide glossary.
700K+
assets documented
in 14 days, across 50 enterprise customers
Auto-generates plain-English documentation for tables and columns by learning from data lineage and how the data is actually used.
Stack
Chosen because they hold up in production, not because they demo well.
Selected Work
Grounded text-to-SQL
Ask any transactional CSV a question in plain English. Every answer carries the SQL that produced it — or a refusal naming exactly what the data would need.
Full-stack AI
A full-stack AI Gmail client where a Python LangGraph agent drives the React UI via natural language — composing, navigating, filtering, and reading email.
Voice-first canvas
A voice-first thinking canvas that turns spoken decisions, plans, and problems into a live typed graph on tldraw.
Real-time voice agent
Real-time AI voice receptionist for clinics — callers talk naturally by phone or browser; the agent books appointments and sends confirmations.
AI agent for API validation
A LangGraph AI agent that validates API endpoint executability by fanning out Claude-powered ReAct agents per endpoint.
Self-correcting media agents
Two small, self-correcting media-generation agents built on Gemini + Google ADK — one for brand mockups, one for narrated explainer videos.
Science, explained
Turns any scientific paper or webpage into an immersive auto-playing documentary — with AI-generated visuals, narration, and live voice Q&A.
Doc intelligence
AI system for logistics document intelligence in TMS workflows — upload docs, ask grounded questions, run structured extraction with confidence and guardrails.
AI video meetings
AI-powered video meeting app for real-time voice conversations with custom AI agents — with live transcription and post-call summaries.
Menu → photography
Turns restaurant menus into stunning, professional food photographs using Gemini + Imagen 4.
Hybrid AI notification routing
A hybrid AI system that decides notify / digest / mute for every incoming message — an LLM (Gemini) makes the call, backed by a deterministic Python signal layer that supplies verified context.
About
I like taking prototypes from “look, it demos” to “look, it’s reliable.”
I'm an AI / software engineer with 2+ years shipping production multi-agent systems — LangGraph orchestration, RAG, agentic harnesses, and evals — that solve real problems for real users.
At Atlan I built support automation, an AI SDLC platform, metrics-glossary agents, and documentation agents that ran at tenant scale with measurable business impact: deflected tickets, saved hours, and shipped documentation for 700K+ data assets.
The interesting part was never the model call. It's the eval loops, the tool boundaries, and the boring plumbing that decide whether an agent is usable on a Tuesday afternoon when something upstream breaks.
Contact
Open to AI-native product teams, and to interesting problems generally. I read everything that lands here.
