Fahmid Uddin

Forward-deployed product engineer

Toronto, Canada

Fahmid Uddin in front of a dark architectural facade

Diagnose / Embed / Build / Deploy

Forward-Deployed Product Engineer — Building and Shipping AI Systems with Teams.

Experience in practice

From the first line
to real-world use.

Hands-on coding
15 years

Career / frontend to AI systems

Engineers led
3–5

GPT Protocol / Airas

Lower RAG cost / faster
~15× / ~20×

LegixAI

Deployment index

Selected work.

Embedded product delivery, a founder-built travel product, and an AI control plane used in Fahmid’s own work.

01AI Account ExecutiveFractional Product EngineerEmbedded with the team as a fractional product engineer, translating ambiguous product and technical needs into delivered releases and continued iteration.Embedded client delivery02OptiflyFounder, Product & GrowthBuilt an AI-native travel planning and booking-intent product that turns a prompt into a live, shareable itinerary, then instrumented the product and acquisition loop.Live product + acquisition loop03Nexore / xOSFounder & AI Systems BuilderBuilding a self-hosted AI control plane used inside Fahmid’s own real work to keep memory, tools, and execution context attached across multiple product and client lanes.Internal control plane / current direction

Primary case file / 2025–Present

AI Account Executive

Fractional Product Engineer

Embedded with the team as a fractional product engineer, translating ambiguous product and technical needs into delivered releases and continued iteration.

Visit AI Account Executive

Diagnose

Clarify product and operating needs

Worked with stakeholders to turn incomplete requirements and changing priorities into a clear delivery path.

Embed

Stay close to the decisions

Operated as an embedded member of the team, connecting product judgment, technical tradeoffs, and day-to-day delivery.

Build

Turn direction into working software

Stayed hands-on through implementation, helping move the product from ambiguity into dependable shipped releases.

Deploy / Iterate

Keep improving after launch

Supported deployment and continued iteration so feedback from real use could shape the next decisions.

Case file / 2025–Present

Optifly

Founder, Product & Growth

Built an AI-native travel planning and booking-intent product that turns a prompt into a live, shareable itinerary, then instrumented the product and acquisition loop.

Visit Optifly
Share-landing views
11,270

Verified 30-day signal

Itinerary loads
13,362

Verified 30-day signal

Booking-click events
201

Verified 30-day signal

Request-to-plan progression
90%

Verified 30-day signal

Optifly workspace showing chat, travel results, and a day-by-day planner
Optifly / Chat, travel options, and itinerary planningAugust 2026

Diagnose

Connect planning to action

The useful product was not another static itinerary. It needed to connect a traveler’s intent to a structured plan, live inventory, sharing, and partner booking rails.

Embed

Work the product and growth loop

Built the customer experience while instrumenting behavior with product and event analytics and reviewing acquisition data at campaign and ad-group level.

Build

Prompt to shareable itinerary

Shipped AI-guided plans with flights, stays, routing, experiences, live inventory, and Expedia / Viator partner click-through paths.

Deploy / Learn

Measure the full path

Used product and campaign signals together to understand request completion, itinerary use, sharing, booking intent, and which acquisition variants earned attention.

Product

  • AI itinerary generation
  • Live travel inventory
  • Shareable plan surfaces
  • Partner booking intent

Measurement

  • PostHog product analytics
  • Operational event visibility
  • Plan progression tracking
  • Booking-click events

Acquisition evidence

  • 167K+ TikTok impressions
  • 29.4K clicks
  • 17.6% CTR
  • ~$0.024 CPC

The TikTok acquisition snapshot covers the same verified 30-day reporting window.

Case file / 2025–Present

Nexore / xOS

Founder & AI Systems Builder

Building a self-hosted AI control plane used inside Fahmid’s own real work to keep memory, tools, and execution context attached across multiple product and client lanes.

Visit Nexore
Internal operating modelHuman-directed control plane
01 / DirectionHuman operator

Sets intent, constraints, and final judgment.

02 / Control planexOS

Routes memory, tools, workflows, and execution context.

03AKnowledge

Semantic memory, Notion, Google Workspace.

03BInfrastructure

AWS Bedrock / Secrets and GCP / Vertex AI.

Current directionAgentCore / Strands

Exploring the next stage of the system.

Diagnose

Stop rebuilding context

Parallel work loses momentum when decisions, notes, and system state are scattered. xOS starts from the operating problem of keeping context durable across real work lanes.

Embed

Use it in the work it coordinates

The control plane is exercised inside Fahmid’s own product, client, campaign, analytics, and launch work rather than positioned as an abstract agent demo.

Build

Connect memory, tools, and execution

The self-hosted system connects semantic memory, agent workflows, Notion, Google Workspace, AWS and GCP services, and communication surfaces around a human-directed loop.

Adopt / Extend

Build on real use

xOS is used in Fahmid’s own operating environment. Bedrock AgentCore and Strands are being explored as the system evolves.

Direction

  • Human sets intent
  • Agents extend throughput
  • Context remains durable
  • Work stays observable

Connected context

  • Semantic memory
  • Notion knowledge
  • Google Workspace
  • Communication surfaces

Infrastructure

  • AWS Bedrock and Secrets
  • GCP / Vertex AI
  • Bedrock AgentCore direction
  • Strands direction

Built for Fahmid’s own work. AgentCore and Strands remain areas of exploration.

Field notes

Stay close to the work.

A few principles shaped by building with teams and staying involved after the first release.

Fahmid Uddin outdoors beside a railing
Portrait / TorontoPersonal archive / 2026

FN / 01

Understand the work before the software

Useful requirements emerge from the real workflow: the handoff that breaks, the decision that lacks context, or the repeated step nobody has named yet.

FN / 02

Make the whole product usable

An AI feature is only part of the job. The way people review its output, recover from mistakes, and get support determines whether it becomes useful day to day.

FN / 03

Keep learning after release

Analytics, traces, evaluations, and error reports show what happens in real use. They help a team decide what to improve next, rather than treating launch as the finish line.

Profile / 2010–Present

A builder, since 2010.

From games and interfaces to client delivery, AI systems, and technical leadership.

Fahmid works across product, code, and infrastructure—staying involved from the first decisions through deployment and iteration.

TypeScript, React, Next.js, and Node.js form the foundation. Recent AI work spans Python, AWS, GCP, Terraform, MCP, RAG, evaluations, and production monitoring.

Technical leadership
Led 3–5 engineers across GPT Protocol / Airas work.
Client delivery
Direct client-site collaboration through Decise.
Current base
Toronto, Canada.

  1. 2010–2012

    Game prototypes and public builds

    Started coding and shipping small game experiments, tutorials, and working demos.
    Early game prototype video artifact
    Archive / 2010–2012

  2. 2013–2015

    3D motion and visual systems

    Built 3D animation and interface experiments that established a durable product-design instinct.
    Early PacMan 3D animation artifact
    Archive / 2013–2015

  3. 2015–2020

    UI, product, and game experiments

    Expanded from visual studies into interface systems, prototypes, and end-to-end product work.
    Collage of early interface and game experiments
    Archive / 2015–2020

  4. 2020–2021

    Metagood / OnChainMonkey

    Joined Metagood as its first engineering hire and helped turn early product direction into shipped foundations.
    Metagood and OnChainMonkey project marks
    Archive / 2020–2021

  5. 2021–2022

    Decise enterprise delivery

    Built software for City BBQ, Pieology, and Chipotle with direct client-site collaboration and production delivery.
    Decise, City BBQ, Pieology, and Chipotle project marks
    Archive / 2021–2022

  6. 2022–2023

    LegixAI workflows and RAG

    Built AI accounting workflows; verified RAG work reduced cost by roughly 15× and ran roughly 20× faster.
    LegixAI product interface
    Archive / 2022–2023

  7. 2023–2025

    AI systems and technical leadership

    Built AI-native products and led teams of 3–5 engineers across GPT Protocol / Airas work.
    AI Account Executive brand mark
    Embedded product delivery

  8. 2025–Present

    Forward-deployed product systems

    Working across AI Account Executive, Optifly, and Nexore / xOS from operating problem through deployment and iteration.
    Optifly workspace showing chat, travel results, and itinerary planning
    Optifly / Travel planning

Direct answers

A little more context.

The role, the work, and how to get in touch.

01What does Fahmid Uddin do?

Fahmid is a Toronto-based forward-deployed product engineer who diagnoses operating problems, embeds with teams, and builds and ships AI systems through deployment and iteration.

02What does forward-deployed mean in Fahmid’s work?

It means staying close to the real workflow and stakeholders: moving from discovery and system design into implementation, deployment, instrumentation, adoption, and ongoing iteration.

03What was Fahmid’s role with AI Account Executive?

As a fractional product engineer, Fahmid embedded with the team and helped move the product from ambiguity through delivery, deployment, and continued iteration.

04Is xOS a customer SaaS product?

No. xOS is Fahmid’s self-hosted AI control plane used in his own real product and client work. Bedrock AgentCore and Strands are current directions, not claims of mature customer production deployment.

05What evidence is available for Optifly?

Verified 30-day signals include 11,270 share-landing views, 13,362 itinerary loads, 201 booking-click events, and 90% progression from request start to plan completion. The acquisition rollup included 167K+ TikTok impressions and 29.4K clicks.

06Where is Fahmid based and how can I contact him?

Fahmid is based in Toronto, Canada. For project enquiries, request a conversation through Applied by Fahmid at fahmid.com/studio/request. For engineering roles and collaborations, connect with him on LinkedIn.

Let’s talk about the work.

Have a project in mind? Start with a conversation. For engineering roles and collaborations, connect on LinkedIn.

Request a project callConnect on LinkedIn