ES
AI Engineer · Innovation & Automation at MiQ Digital

Ethan Sam

I build LLM agents and automation systems that ship and stick — the kind that actually get used in production.

Daily Briefing Buddymost-used internal AI tool in MiQ history

93%
US Adoption
46K+
Conversations
1,400
Person Org
Resume
Inside the agent

Anatomy of a daily briefing

A replayed production trace from this morning's run. Parallel tool calls fan out, a reasoning step synthesizes, and the result lands in a Slack DM — all under 1.3 seconds, every weekday at 08:00.

Tracing
Daily Briefing Buddy
trace--------
duration—
servicedaily-briefing-buddyenvprodversion2.4.1userusr_a3f7c2
08:00:00.000cron.weekday(08:00 EST)trigger
08:00:00.000├tools.parallel4 calls · fan-out
08:00:00.000├GET/gcal/v3/events?date=today142 ms·
08:00:00.000├SOQLsfdc.Campaign__c WHERE Status='Live'287 ms·
08:00:00.000├SQLdbx.ml.anomaly_scores (7d)401 ms·
08:00:00.000└GET/slack/conversations.history198 ms·
08:00:00.401├reasoning.synthesize784 ms·
08:00:01.185└POST/slack/chat.postMessage67 ms·
#briefings·DM with you
Daily Briefing BuddyAPP8:00 AM

Sample trace · client names and identifiers anonymized.

Selected work

9:41

Anchor

Build better habits, one day at a time

Edit

Exercise (20 Pushups)

6 times a day

92
🔥6
Fr
Sa
Su
Mo
Tu
We
Th
TODAY

30 minutes of Audiobook

Daily

78
🔥3
Fr
Sa
Su
Mo
Tu
We
Th
TODAY

Morning Meditation

Daily

85
🔥1
Fr
Sa
Su
Mo
Tu
We
Th
TODAY
Habits
Friends
Messages
Settings
View Live App

Anchor

Build better habits with friends

iOS App Store

Shipped to production with Capacitor

On-device ML

TensorFlow.js inference for streak insights

26 RLS policies

Every table locked down, zero exposed endpoints

Real-time sync

Live messaging and friend activity via Supabase

React 19SupabaseCapacitorTensorFlow.jsFramer Motion
SourceLive App
GPT-2 · Layer 6 attention
Induction head

NeuroScope

Building a browser-based mechanistic interpretability toolkit for transformer models — logit lens projections, steering vector injection, QK attribution decomposition, and 3D PCA trajectories of the residual stream. Active research collaboration with a CMU researcher; aim is to make interpretability work accessible without a Python environment or GPU.

TypeScriptWebGPUTransformers.jsTransformerLensReactVite

Browser-native · No GPU required · Research-stage

#campaign-alerts
Insights Botjust now

CTR for Campaign #4421 dropped 47% vs 7-day baseline.

Z = −3.2GPT insight ↗
Z-score · GPT

Campaign Insights Bot

Built before the interview, not after. Designed an AI-powered Slack monitoring system with Z-score anomaly detection and GPT-generated campaign insights. Closed-loop feedback lets the model improve from analyst corrections in real time.

ReactRechartsOpenAI GPTTailwind CSSVercel

Built proactively — secured the role before the first interview

SourceDemo
Kalshi · Daily HIGH temp
EnsembleMarket
+Δ Edge

Weather Edge

A quantitative trading system that finds mispriced daily high temperature brackets on Kalshi prediction markets by fusing 194 AI and physics-based weather model forecasts against real-time market prices.

PythonasyncioKalshi APIOpen-MeteoKDE (scipy)Claude AI

194 ensemble forecasts | 7 pre-trade risk guards | Live system

Demo

Experience

MiQ Digital

AI Engineer, Innovation & Automation

Jul 2025 — Present

New York, NY

  • ◆Daily Briefing Buddy — built and shipped an AI agent end-to-end across a 1,400-person organization. 93% US adoption, 46,000+ conversations, leadership-named the most-used internal AI tool in company history.
  • ◆Glean platform technical lead for US Commercial Ops — ran live demos and cross-functional training while shipping 5+ production agents across sales enablement, deal desk, and onboarding.
  • ◆Architected the €5.7M “Digital Cash Out” implementation plan for a global pharmaceutical client — phased rollout across 35 markets, adopted by client leadership over the original higher-risk approach.
  • ◆Drove the Ad Ops Automation initiative — Claude API + DSP API integrations (REST + webhooks) cutting tag sheet handling time 70–85% across hundreds of monthly campaigns ($900K+ annualized savings).
  • ◆Won 1st place at MiQ’s Retool hackathon as technical lead — Servicing Hub campaign lifecycle platform integrating Salesforce, Databricks, and Jira APIs, projected to save 20,000+ hours annually.

Lyminal LLC

Founder

2024 — Present

New York, NY

  • ◆Software studio shipping AI-powered products and applied ML research. Multiple ventures in active development.
  • ◆Anchor — built and shipped a social habit-tracking iOS app applying behavioral science principles (habit stacking, identity-based goal framing) with real-time accountability, group streaks, and push-driven engagement loops.
  • ◆NeuroScope-Web — building a browser-based mechanistic interpretability toolkit with a CMU collaborator: logit lens, steering vectors, QK attribution, and 3D PCA trajectories rendered in TypeScript/WebGPU.

NYU Langone Medical Center

AI & Cybersecurity Intern

May 2024 — Aug 2024

New York, NY

  • ◆Built an AI-driven chatbot to automate GRC and vulnerability management workflows — reduced triage time for the security operations team.
  • ◆Led third-party vendor risk assessments ensuring compliance with HIPAA, PCI-DSS, and SOC 2 frameworks.
  • ◆Conducted vulnerability assessments to validate patch deployment and reduce exposure across enterprise systems.

NYC Cyber Command

Cybersecurity Auditor

May 2021 — Aug 2023

New York, NY

  • ◆Administered ICS security assessments across 50+ mission-critical OT/ICS systems for the City of New York using NIST CSF and SP-800-53.
  • ◆Coordinated remediation across 10+ city agencies — 20% reduction in non-compliance, 100+ vulnerabilities remediated.
  • ◆Authored the internal Audit Guide, reducing audit time by 20%.

Recognition

1st Place · MiQ Retool HackathonServicing Hub · technical lead · 2026
Best Challenge · NYU CSAWCNN-based phishing detection · 2023
Black Hat Briefings ScholarAwarded for early cybersecurity work · 2023

Education

New York University

M.S. Cybersecurity

Aug 2022 — Jun 2025

GPA: 3.96

NSA-CAE DefenseNational Centers of Academic Excellence in Cybersecurity (NCAE-C)
Exploring Adversarial Machine LearningNVIDIA Deep Learning Institute
  • ◆Key Coursework: Network Security, Digital Forensics, Cloud Security, Applied Cryptography
  • ◆Co-developed CNN-based phishing detection system at CSAW (Best Challenge Award) — attracted 300+ participants, scholarship to Black Hat Briefings

CUNY Hunter College

B.A. Computer Science & Applied Psychology

Aug 2017 — Jun 2021

GPA: 3.76

  • ◆Led CodePath.org cybersecurity program with 70% placement rate
  • ◆Enhanced threat detection accuracy by 20% using IBM QRadar
  • ◆Black Hat Briefings Scholarship for early cybersecurity work

Ideas

I publish when I learn something worth sharing.

Build Log8 min read

How I Built an AI Agent That 93% of My Company Actually Uses

Most enterprise AI tools die the same death: a flashy launch, then silence. Here's how a push-based design pattern broke that cycle — 46,000 conversations and counting.

Read
View all ideas

Get in touch

Open to AI engineering roles where adoption is the metric — applied AI, agent infrastructure, and AI safety work. Equally happy to talk about interpretability research or applied behavioral science in product.

Fastest reply: email or LinkedIn. Quietly hiring? I'm listening.

Get In TouchResume