Business Planner · AI workflow builder · Dallas, TX

Quandong Qian

I turn messy workflows across SAP, SQL, Excel, and internal tools into reliable automations with dry runs, human review, and explicit failure boundaries. I am now exploring AI FDE / Solutions / Operations Engineer roles.

  • 13 groups of TI planners use my automations and reports every day
  • 700+ members in Dallas's first Chinese AI community (founder); 246 attendees across five events
  • 300 members in an internal AI community with biweekly sessions
Quandong Qian
Dallas, TX

Experience

From supply-chain problems to controlled AI workflows.

I start with the operating outcome: map the workflow, exceptions, and ownership boundary, then choose SQL, scripting, browser automation, or an LLM.

Texas InstrumentsDallas, TX
Business Planner, Supply Chain Apr 2026 – Present
  • Automated the recurring planning reports and reconciliation checks, saving about 5 hours a week of manager and director review time; taught planners in 13 groups to run the tools themselves, saving each about 2 hours a week.
  • Scripted SAP GUI end to end with Python (win32com) for sales-order scheduling, correction, and release; it runs in an isolated session with a dry-run default, so a bad run cannot touch live orders.
  • Built Selenium form-fillers for 12+ internal request types with a human-review stop before submit; nothing auto-submits.
  • Replaced manual SAP / Spotfire / Excel cross-referencing with direct SQL on a data-virtualization layer, turning supply-gap reports, the weekly risk email, and quarterly review decks into generated outputs.
  • Founded and run a 300-member internal AI community; present the automation playbooks every two weeks so planners can reuse them.
Industrial Engineer, Sherman fab Aug 2023 – Apr 2026
  • Cut reticle wait time across the fab by 80% by automating reticle scheduling and handling.
  • Wrote the real-time dispatching (RTD) logic for the new Sherman fab from zero as part of the build-out team, balancing throughput, cost, cycle time, and on-time delivery.
  • Designed the lithography scheduling logic, adding load-port and expression-based constraints the previous logic missed.
  • Shipped a Vectara + Webex knowledge-base chatbot so night-shift operators resolve MES issues from step-by-step guidance instead of waiting for day-shift support.
  • Redesigned MES message structures and reports for integrated load/unload, reconciling data mappings across 4,802 variations.
  • Automated daily dispatch metrics, reticle-delivery ETL, and version checks; migrated dispatch jobs across redundant servers.
Dallas Chinese AI Community · Real Agent Use CasesFounder & Lead Organizer (side project) · 2026 – Present
  • Founded and grew Dallas's first Chinese AI community to 700+ members: 5 live events from Apr to Aug 2026, 246 cumulative attendees, 70 at the largest.
  • Ran hands-on workshops (40 and 59 attendees) where nontechnical participants built and ran a working email daily-brief agent from a template, then demoed their own automations; also organized a 37-person AI hackathon.
  • Drew a cross-industry room: programmers, data analysts, bank staff, accountants, lawyers, real estate agents, restaurant and small-business owners. Aug 22 registration survey (all 59 guests): 66% working professionals, 49% already using AI daily.
  • Designed and taught Agent 101, a bilingual course and hands-on workshop; attendees build a working Gmail agent live from a template. View the English lecture slides →
  • Shipped realagentusecases.com, the community's bilingual content site and course home.
Texas A&M UniversityClass of 2022
  • B.S. Electrical Engineering, College Station, TX.

Selected case studies

Four examples of how I work as an FDE.

These span a live AI receptionist for outside businesses, a scheduling system an outside business runs every week, internal enterprise delivery, and hands-on deployment with nontechnical users. Each starts with a concrete problem and keeps human judgment and failure boundaries explicit.

01 · AI receptionist · LiveCall or text (469) 425-4142 · English, Spanish, Chinese

AI receptionist that answers by voice and SMS, checks availability, and books

Problem

Small businesses miss calls and texts while serving customers, every booking turns into back-and-forth over times, and prospects should be able to try the receptionist before a sales call, with no signup.

Delivered

An OpenAI Realtime voice line and an SMS agent share one tool set (check availability, create / find / reschedule / cancel booking) over a booking API on Cloudflare Pages Functions + Supabase Postgres. Every confirmed change texts both the customer and the owner. Public landing page and self-serve booking page at aimanjack.com; the same SMS pattern runs on a DFW car dealership's number.

Guardrails

Quotes only approved prices and hands off everything else; double-booking is blocked in the database, not just in the prompt; reschedule and cancel follow a notice window; never takes payment; every write produces a confirmation text, so a missed conversation still surfaces.

Architecture & trade-offs

Twilio is a pure pipe: TwiML hands the call to a WebSocket bridge and the caller's number rides along as a parameter, so voice and SMS reach the same tool executor. One booking API is the single write path, and capacity checks plus advisory locks live in Postgres functions, so the worst a model mistake can do is say the wrong thing, never create a bad booking. A debug endpoint exercises each tool without placing a call, and a smoke test runs against the API after every change.

02 · Ops automation · In productionSaves the ops lead ~4 hours a week

Weekly host scheduling system for a TikTok Shop live-commerce agency

Problem

Every week the ops lead collected hosts' and operators' availability by chat, hand-dragged the schedule across several live rooms in a Lark sheet, and hand-wrote a per-host plan every day. Conflicts between the schedule and people's real availability were caught by eye.

Delivered

Hosts submit next-week availability through signed per-person links. A Cloudflare Worker (KV + cron) reads scores, pair compatibility, hour limits, and room requirements from Lark Sheets, runs a constraint solver, and writes the draft schedule plus an explicit gap list back to Lark. An admin panel shows who has submitted and the state of the draft.

Guardrails

The system only drafts; the ops lead reviews and publishes. Sessions are never shorter than two hours; slots nobody can cover are left empty and listed as gaps instead of force-filled. No personal Lark identities, no messages sent by the machine, secrets only in Cloudflare.

Architecture & trade-offs

Runs entirely on Cloudflare (Worker + KV + cron) with Lark app credentials, so it depends on nobody's laptop. Lark stays the system of record: submissions land in a Lark Base, the draft lands in a Lark sheet, and the ops lead keeps editing where she already works. Each submission link carries an HMAC signature for the person and the week, so stale links cannot submit. 21 regression tests cover the solver and link signing, CI/CD deploys on push, and a handoff doc lets the next maintainer run it without me.

03 · Internal enterprise · Daily useIn daily use by planners across 13 groups

Supply-chain operations automation at Texas Instruments

Problem

Scheduling, sales-order corrections, supply gaps, and recurring reviews required repeated reconciliation across SAP, Spotfire, Excel, and internal forms.

Delivered

Python against an isolated SAP GUI session, direct SQL to the data-virtualization layer, Selenium for 12+ form types, and reports, email, and review decks generated from one data source.

Guardrails

Dry-run by default; forms are filled but never submitted; exceptions go to a person; the automation session never disrupts live user work.

04 · Live users · Bilingual delivery5 events · 246 attendees

Agent 101: helping nontechnical users build a first agent live

Problem

The room spans banking, accounting, law, real estate, restaurants, and software, with a wide range of AI experience. One workshop has to get everyone building.

Delivered

A bilingual course and Gmail-agent template that connects concepts, a live demo, a hands-on build, and one-on-one troubleshooting.

Outcome

Founded Dallas's first Chinese AI community and grew it to 700+ members; 246 cumulative attendees across five events, 70 at the largest, and 59 registration responses for the Aug 22 session.

More open-source projects and experiments
Personal agent · Daily driver
Gmail → P0 / P1 / P2 action list

Turns the previous 24 hours of email into prioritized tasks with time windows and source links; read-only end to end and saves about 25 minutes a day.

Content pipeline · Gated publishing
SEO / GEO publishing pipeline

Skips a run when sources are weak, requires a named source for every claim, and turned 22 deleted low-quality pairs into permanent agent rules.

Client demo · 9 custom pages
Local-business audit and outreach agent

Builds tailored demos from public information and routes outreach to Gmail drafts for human review.

Video pipeline · 7 episodes
Kids' AI education channel

Qwen3-TTS narration, code-rendered frames, ffmpeg assembly, loudness checks, and YouTube API upload with failure-safe queue state.

Agent · SMS receptionist · Deployed
Used-car dealership SMS AI receptionist
  • Handles inbound texts for a dealership: collects the request, offers open appointment slots, writes confirmed bookings to a database, and texts the owner a summary.
  • OpenAI function calling on a Cloudflare Worker with D1; Twilio stays a pure SMS pipe. Ships a direct test endpoint and redeploys for a new store by swapping one config file.
OpenAI APICloudflare D1Twilio
Agent skill · Open source
cheap-flights
  • Tries nearby airports, flexible dates, and split one-ways automatically; compares bag-inclusive totals.
  • Hands back links, never books; fails openly instead of inventing a quote.
CLIAgent skill
Open source · 6 workflows
Obsidian AI content OS
  • Open-sourced 6 workflows: AI-news intake, morning briefing, favorites-to-knowledge-cards (yt-dlp + Whisper), link ingest, 90-second scripts, email triage.
  • Closed the loop: information in, AI sorts, a human filters, content out.
ObsidianWhisperyt-dlp
Open source · Agent skill
track-ai-visibility
  • Audits brand visibility in AI search: prompt coverage, competitor mentions, cited domains, content gaps.
  • Saves evidence as JSON and Markdown; marks unobservable platforms unobserved, never fakes results.
CodexClaude Code
Open source · CRM
booking-research-agent
  • Researches the contact and company on each new booking, writes to a Notion CRM, drafts a prep email in the owner's voice.
  • Drafts only; never sends, never commits price or scope.
Notionn8n

Community

Taking AI off the screen and into a real room.

I founded and lead Dallas's first Chinese AI community, now 700+ members strong: programmers, data professionals, bankers, restaurant and small-business owners, accountants, lawyers, and real estate agents, all here to put AI to work.

Every month, one live event in Dallas. Hands-on demos, real workflows, and the kind of conversations that only happen in the room.

700+
Community size
5
Live events
246
Cumulative attendees

Events

Monthly in Dallas.

Live sessions in Dallas: hands-on workshops, real agent walkthroughs, and hackathons, in Chinese and English. All events run through Luma.

2026.09.24AI Front Desk 101 for Small Clinics (free workshop for DFW practice owners, with Amanda Li)Upcoming · Register →
2026.08.22Build Your First AI Project — No Experience Required (free workshop for students and parents)59 attended
2026.07.18AI Workflow Automation in Practice — Session 340 attended
2026.05.17OpenClaw: Enter the World of AI Agents — Meetup #270 attended
2026.04.18Dallas AI Hackathon37 attended
2026.04.02Dallas Openclaw40 attended


About

The problem I want to solve.

Many enterprise AI projects stall at the demo because the model is only one part of the work. The rest is workflow, permissions, exceptions, ownership, and user behavior. I work from those constraints and connect the useful part of AI to the job people already need to finish.

I currently work in supply-chain Business Planning at Texas Instruments. Everything outside the day job — the community, the course, the open-source workflows, the demos for local businesses — is practice at the same question: how AI lands in real work. They are training grounds, not a second career I plan to run.

I am looking for a full-time AI FDE / Solutions Engineer role, and I am glad to talk enterprise workflow automation, customer demos, and AI user education.

Side projectaimanjack.com, AI demos and websites for local businesses
Based inDallas, Texas
Day jobBusiness Planner (supply chain) at Texas Instruments
CommunityFounder of Dallas's first Chinese AI community, now 700+ members
GitHubgithub.com/qianquandong
TargetAI FDE · Solutions / Operations Engineer · Enterprise AI workflow

If you need an AI FDE who can enter the workflow and ship the system, let's talk.