Founding Engineer · TrueHorizon AI

Deepankar Bhadrasen

Deepankar Bhadrasen

Scaling output without scaling headcount. I build end-to-end AI automation that ships and runs in production for enterprise teams.

Work with Me See the work  

$2.5M ARR at TrueHorizon AI20+ AI agents in productionFortune 500 clients

01 About

I build AI systems that hold up in production.

Most of my work is for Fortune 500 teams: AI agents and end-to-end automation that take over the slow, manual parts of sales, finance and operations, wired into the tools those teams already run on.

As Founding Engineer at TrueHorizon AI, I worked across the core automation infrastructure, client delivery and the repeatable frameworks that helped scale the company to $2.5M ARR.

Right now I am also building THAI UI, TrueHorizon AI's AI agent platform, with Claude Code and Codex.

Read the full story
ARR at TrueHorizon AI
$2.5M
Helped scale it as Founding Engineer
AI agents
20+
Deployed and running in production
End-to-end systems
10+
Designed, built and deployed
Client base
Fortune 500
Where most of these systems run
02 Speaking

Showing the work, live.

At the n8n Builders Challenge in Ho Chi Minh City, I presented live n8n workflows on the big screen to a room of builders, then worked hands-on with the teams through the build hours.

Deepankar presenting to a room of builders at the n8n Builders Challenge, a consultation intake form on the big screen behind him
Speaking at the n8n Builders Challenge · Ho Chi Minh City, Sep 2026
Event
n8n Builders Challenge
City
Ho Chi Minh City, Vietnam
Date
Saturday, 19 Sep 2026
Format
Live workflow demos, then hands-on build hours

Before that, I was a guest consultant at an off-site hosted by a Fortune 500 company, showing GTM leaders how to implement AI practically and speed up adoption.

Invite me to speak
Deepankar pointing at an n8n workflow on screen beside the n8n Builders Challenge banner
Walking through a live n8n workflow
Gesturing at a screen showing a consultation intake form
Breaking down an intake flow on screen
Deepankar at a laptop, hand on chin, thinking through a problem
Thinking through a build
Wide view of the busy build room, Deepankar typing at the right
The build room, mid-challenge
03 Case Studies

Systems that ship.

A selection of systems built for enterprise and Fortune 500 clients. Each canvas maps how the system runs, from the first trigger to the final record.

Canvases and run logs are illustrative and contain no client data.

01Revenue Automation

Automated Sales Pipeline

ClientUS-based nonprofit · Catholic organization

Problem
Outreach was manual: every lead was qualified, followed up on and entered into the CRM by hand, for hours every day.
System
An end-to-end pipeline that replaces manual outreach with intelligent workflows. Leads are qualified automatically, follow-ups sequenced and all data synced to Salesforce, with no human input at any step.
Result
Eliminated hours of daily manual outreach. Scaled lead engagement without adding a single person.

Diagram of how the Automated Sales Pipeline runs: a new lead triggers the workflow, OpenAI qualifies it, it is scored and routed, follow-ups are sequenced and activity is logged to Supabase, then the record syncs to Salesforce.

  • n8n
  • Salesforce
  • OpenAI
  • Supabase
Build something similar
Lead pipeline How it runs Production

Run logillustrativeSucceeded

  1. Lead captured
  2. Qualified by AI
  3. Scored and routed
  4. Follow-ups sequenced, activity logged
  5. Synced to Salesforce

02Process Automation

Call Grading System

ClientEnterprise · Sales and support teams

Problem
Coaching depended on hours of manual call review across sales and support teams.
System
Automatically ingests, transcribes and scores calls against custom rubrics. It surfaces coachable moments and flags underperforming patterns, all before the manager opens their inbox.
Result
Replaced hours of manual call review. Enabled consistent, data-driven coaching at scale.

Diagram of how the Call Grading System runs: a new call recording arrives from Fireflies, the transcript is fetched, OpenAI scores it against the rubric and flags coachable moments, then scores are stored in Supabase and a summary goes to the manager.

  • Fireflies API
  • OpenAI
  • n8n
  • Supabase
Build something similar
Call grading How it runs Production

Run logillustrativeSucceeded

  1. Call recorded
  2. Transcript ready
  3. Scored against rubric
  4. Coachable moments flagged
  5. Scores stored, summary sent

03Document Intelligence

Invoice Parsing System

ClientE-commerce brands

Problem
Invoices in every format, keyed in by hand across high-volume order flows.
System
Reads incoming invoices in any format, extracts structured data, validates it against order records and writes clean entries to the database. Thousands of documents, no manual input required.
Result
Eliminated manual invoice entry entirely. Scaled to high document volumes with zero additional effort.

Diagram of how the Invoice Parsing System runs: an invoice arrives, the file is read in any format, OpenAI extracts the fields, they are validated against order records in MongoDB, and a clean entry is written to the database.

  • OpenAI
  • MongoDB
  • n8n
Build something similar
Invoice parsing How it runs Production

Run logillustrativeSucceeded

  1. Invoice received
  2. Document read
  3. Fields extracted
  4. Matched to order record
  5. Clean entry written

04Financial Intelligence

Finance Document Agent

ClientEnterprise · Financial services

Problem
Loan applications were assembled by hand from unstructured financial documents, the slowest part of every review.
System
An AI agent reads and interprets unstructured financial documents using vector search and LLMs, then populates and formats a complete loan application.
Result
Drastically cut loan application processing time. Removed manual document review from standard application workflows.

Diagram of how the Finance Document Agent runs: documents arrive, an AI agent reads them with Gemini, retrieves context through Pinecone vector search, reasons with an OpenAI model, and saves a formatted loan application to Supabase.

  • OpenAI
  • Pinecone
  • Supabase
  • Gemini
Build something similar
Loan document agent How it runs Production

Run logillustrativeSucceeded

  1. Documents received
  2. Agent started
  3. Documents read
  4. Context retrieved
  5. Application drafted
  6. Formatted application saved
  1. 01 Automated Sales Pipeline
  2. 02 Call Grading System
  3. 03 Invoice Parsing System
  4. 04 Finance Document Agent

The four above are a sample of 10+ end-to-end systems and 20+ AI agents deployed in production, mostly for Fortune 500 clients across sales, finance and operations.

04 Track Record

Every role, and what it delivered.

Where I have built, advised and presented, with the outcome of each.

  1. Early 2025to present

    Founding Engineer

    TrueHorizon AI

    One of the first members of the team. Built the core automation infrastructure that enterprise client delivery runs on, and the repeatable frameworks behind it.

    Architecture · Client delivery · Repeatable frameworks

    $2.5M ARRHelped scale the company

  2. Nowin active development

    AI Engineer

    THAI UI · TrueHorizon AI

    Designing the AI agents, skills and workflow automation inside THAI UI, TrueHorizon AI's AI agent platform, and developing the product around them myself: the Vue interface and the Node and MongoDB services underneath. Built with Claude Code and Codex.

    AI agents · Workflow automation · Product development

    End to endFrom agent logic to the product UI

  3. Guestengagement

    Guest Consultant

    Fortune 500 off-site

    Brought in to show GTM leaders a practical path to AI adoption across their teams.

    GTM leadership · AI adoption · Practical implementation

    Fortune 500Host of the off-site

  4. Ongoingco-founded venture

    Co-founder

    Qixazow

    Writing, consulting and video on n8n, Claude and business AI, with two fellow AI professionals.

    Publication · Consulting · Video

    Live venturePublication and consultancy

  5. 19 Sep 2026Ho Chi Minh City

    Speaker

    n8n Builders Challenge

    Demoed live n8n workflows to the whole room, then spent the build hours helping teams with their builds.

    Live demos · Workflow design · Build hours

    Big-screen demosThen build hours with the teams

  6. Sincecollege

    AI Automation Engineer

    Independent

    Automation systems for finance, e-commerce and enterprise sales teams, starting with real client work delivered alongside coursework.

    Finance · E-commerce · Enterprise sales

    First clientPaid, before graduating

What teams hire me for

Three kinds of systems, each backed by a case study above.

Revenue Automation

AI-powered sales pipelines, lead qualification systems, and CRM integrations that replace manual outreach and scale revenue without adding headcount.

Case studyAutomated Sales Pipeline

Document Intelligence

Automated parsing and analysis of financial documents, invoices, and structured data, turning unstructured input into clean, actionable output at volume.

Case studiesInvoice Parsing SystemFinance Document Agent

Process & Content Systems

End-to-end content pipelines, call grading systems, and cross-platform workflow automation built to run reliably in production, not just in demos.

Case studyCall Grading System

Industries served

  • FinanceFinance Document Agent
  • E-commerceInvoice Parsing System
  • Enterprise salesCall Grading System
  • NonprofitAutomated Sales Pipeline

Builds with

  • Claude CodeCoding agent
  • CodexCoding agent
  • n8nWorkflow orchestration
  • Full stackVue, Node and MongoDB
05 Thoughts

Notes from the build.

Essays on agents, automation and putting AI to work inside real companies. Each first published on TrueHorizon AI.

Enterprise AI6 min read

Claude Design for Enterprises: 6 Things to Know Before You Prototype

It is not a prettier Canva. Design system extraction, the four-part prompt framework, and an honest map of where it still breaks. Read the essay AI Agents5 min read

The AI Agent That Does YouTube Research While You Sleep

Viral outlier detection, transcript analysis, and thumbnail vision models: a content research pipeline built on validated data instead of guesswork. Read the essay Automation5 min read

Stop Prototyping Slowly: n8n MCP + Claude Code

From plain-English prompt to production-shaped workflow in about 20 minutes. Why MCP changes what Claude actually knows about your stack, and where humans still matter. Read the essay
All writing
06 Contact

Let's build something exceptional.

Ready to automate your most time-consuming workflows? I'm available for new projects, consulting engagements, and speaking opportunities.

Email sales@deepankar.xyz
Location Bhopal, India · Available globally
LinkedIn deepankarbhadrasen
$5,000 $12,000 $50,000+

By sending this, you agree I may use these details to reply to your enquiry. See the privacy policy.