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.

Verified proof

Named work. Scoped evidence.

Hands-on coding
15 years

Career / frontend to AI systems

Engineers led
3–5

GPT Protocol / Airas

Itinerary loads
13,362

Optifly / verified 30-day signal

Lower RAG cost / faster
~15× / ~20×

LegixAI

Deployment index

Three systems, three operating contexts.

The work is grouped by the problem surface, Fahmid’s role inside it, and the evidence that can be attributed to that deployment.

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.

Embedded role

  • Stakeholder proximity
  • Product and technical judgment
  • Hands-on implementation
  • Sustained team collaboration

Delivery arc

  • Clarify ambiguity
  • Shape the delivery path
  • Ship working releases
  • Iterate after deployment

Value created

  • Aligned product decisions
  • Reduced handoff friction
  • Production-minded execution
  • Faster feedback loops

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 public workspace showing chat, live travel results, and a day-by-day planner
Public workspace / Chat + results + plannerLive product page / 12 Aug 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

Exploration, not a mature deployment claim.

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

Keep the claim boundary explicit

xOS is an internal operating system today. Bedrock AgentCore and Strands are active directions under exploration, not claims of mature customer production deployment.

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

Claim boundary: xOS is used in Fahmid’s own operating environment; this portfolio does not present it as a customer SaaS or mature AgentCore deployment.

Field notes

What changes when the engineer stays close.

Working principles drawn from product delivery, client collaboration, instrumentation, and operating the systems after launch.

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

FN / 01

Stay close to the operating surface

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

Ship the operating system around the feature

An AI feature is rarely the whole product. The surrounding operating model, delivery process, and support systems determine whether it can be used and improved.

FN / 03

Instrumentation is part of delivery

Product analytics, traces, evaluations, and error visibility turn a release into an observable operating loop instead of a one-time handoff.

Profile / 2010–Present

Fifteen years across the full product surface.

The chronology moves from games and interface work into client delivery, AI workflows, cloud systems, and forward-deployed product engineering.

TypeScript, React, Next.js, and Node.js form the strongest historical through-line. Recent AI-agent work adds substantial Python, AWS, GCP, Terraform, MCP, RAG, evaluations, traces, 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 public product artifact
    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 public brand mark
    Public brand / 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 public workspace showing chat, live results, and itinerary planning
    Verified product / live workspace

Direct answers

The useful facts, without the pitch deck.

These visible answers are also the source for the site’s FAQ structured data and answer-engine context.

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. The direct contact address is fahmid.me@gmail.com.

Contact / Toronto, Canada

Bring the messy operating problem.

For forward-deployed engineering, applied AI product systems, founding-team work, or client delivery conversations.

fahmid.me@gmail.com