Skip to content
$ SAMIHAROON.COM
BOOK CALL

> CTO & AI ADVISOR — KARACHI

SAMIHAROON

I have taken the co-founder path twice and the staff engineer path through high-volume social, payment, retail, and enterprise systems. My work is usually called in when the problem is expensive to misunderstand: model workflows, payment paths, low-latency rooms, team operating rhythm, or architecture that has to keep holding after launch.

Book 15-min callResume.pdfEmail

> 000 / Current focus

Senior technical judgment for systems where ambiguity is more expensive than code.

  • [01]Production AI workflows
  • [02]Fintech and payment infrastructure
  • [03]Realtime products and moderation
  • [04]Engineering leadership and operating rhythm

STATUS: OPEN_TO — ADVISORY / ARCHITECTURE REVIEW

> 001 / About

Work at the edge of product pressure, architecture, and risk.

My work has moved between founder rooms and enterprise systems: AI/ML products, fintech infrastructure, realtime media, moderation, search, and teams that needed a clearer way to ship.

01 Founder judgment

Architecture decisions before the safety net exists.

Twice I have worked from zero to one, where unclear ownership, weak delivery rhythm, and fuzzy product bets show up immediately in the system.

02 Systems under load

Moderation, search, routing, payments, and realtime rooms.

The work has usually lived where accuracy, latency, cost, and trust all matter at once: social platforms, banking modules, retail analytics, and enterprise routing.

03 Team leadership

The operating system around the code matters.

Hiring, onboarding, review standards, API documentation, architecture notes, and execution cadence decide whether a team can keep quality under pressure.

04 Production AI

Start with the workflow before choosing the model.

A production AI build needs owners, data boundaries, approval paths, fallbacks, and measurement before it needs a larger prompt or a more expensive vendor.

> 002 / Work_history

Selected work across realtime products, payments, search, and production AI.

May 2024 - Present

Vupechat Inc.

Co-Founder and CTO

Built a niche VTuber social platform from zero to one: WebRTC motion plugins, avatar rooms, spatial audio, moderation, matching, and creator workflows.

>> MVP in 3 months, 18,000+ VTubers, 99.2% moderation accuracy, sub-100ms latency targets for 1,000+ concurrent sessions, and an embedding-based matching system.

Dec 2022 - Apr 2024

Blossom.team

Staff Software Engineer

Built ML moderation and search systems for a social product where accuracy, relevance, and review load all had to improve at the same time.

>> 100,000 daily interactions, 90% manual-review reduction, $200K annual savings, 300% search relevance lift, and user satisfaction from 3.2 to 4.8/5.

Feb 2022 - Jun 2023

Remotebase

Technical Lead

Owned architecture and delivery across enterprise client work, including fintech systems, analytics tooling, and cloud migration projects.

>> Coordinated teams up to 25 engineers, managed a 12-person Arrow Payments team, cut AWS costs by 35%, and shipped analytics tooling to 5 enterprise clients.

Dec 2020 - Apr 2022

Afiniti Inc.

Senior Software Engineer

Worked on performance-sensitive C/C++ APIs, shared-memory systems, queue management, and AI-driven omnichannel routing infrastructure.

>> 60% benchmark improvement, 30% memory reduction, high-availability enterprise deployments, and structured API documentation that shortened onboarding.

Jan 2020 - Dec 2020

Kepler Analytics

Software Engineer

Built analytics reporting and integrations for retail operations where reporting speed and metric quality shaped customer value.

>> 25+ global retail chains, 500+ business metrics per client, POS and ETL integrations, and 70% faster data processing.

Nov 2018 - Dec 2019

Oraan

Software Engineer

Took a fintech product from concept to production across PWA, mobile, Java Spring Boot services, identity, and deployment.

>> Owned the GCP roadmap, Keycloak-based access control, secure APIs, PostgreSQL data management, and production delivery.

Oct 2017 - Oct 2018

Techlogix

Software Engineer

Built banking payment modules and infrastructure improvements for enterprise clients with strict security and deployment constraints.

>> $1M+ daily transaction modules for Bank of Punjab with zero security incidents, plus NEJM cluster optimization cutting costs by 25%.

> 003 / Principles

Clear systems are easier to build, challenge, and operate.

01Architecture is a business decision before it is a diagram.

02A useful AI system names the owner, the approval path, and the metric.

03Realtime products fail in the distance between latency, moderation, and trust.

04Engineering standards should survive urgency, not appear only after it.

> 004 / Education

Software engineering, technical communities, and early leadership practice.

Software Engineering

NED University

Bachelor of Engineering

Studied software engineering with a focus on systems, product delivery, and applied engineering foundations.

>> A- / Cum Laude, supported by a fully funded Koshish Foundation scholarship.

> 004.1 / Volunteering

Competitions and Events

ACM NED

Event Chairperson and Director

Led student community programming across technical workshops, competitions, and operating rhythm.

>> Built the early leadership muscle around coordination, standards, and public technical work.

Developer and Business Communities

GDG and GBG Karachi

Executive and Partnerships

Worked across developer and business communities on meetups, workshops, local events, and corporate partnerships.

>> Kept technical community work close to founders, engineers, operators, and the people adopting new tools.

> 005 / Stack

Production range.

Tools I use when the work needs to reach production.

The stack changes by problem. The standard does not: clear boundaries, measured risk, fast recovery, and work a team can keep operating after launch.

> Product and backend systems

APIs, service boundaries, realtime rooms, and integration surfaces.

Used when the product needs clear ownership, predictable contracts, and code that can be reviewed under pressure.

FastAPI / NestJS / ElysiaJS / Spring Boot / Node.js / Bun

> Production AI

Model workflows, local inference, evaluation, and document pipelines.

Used when the model is only one part of the system: prompts, retrieval, review paths, fallbacks, and metrics all have to hold together.

LangChain / vLLM / Google A2A / PyTorch / Claude Code / Ollama / Docling / PaddleOCR

> Cloud and delivery

Deployment paths, queues, traffic routing, observability, and cost control.

Used when the system has to survive release pressure, traffic spikes, failure recovery, and the bill at the end of the month.

AWS / GCP / Docker / Terraform / Kafka / Nginx / Jenkins

> Data and product surfaces

Persistence, search, realtime interfaces, and product surfaces.

Used when users need fast screens, searchable history, clean state, and realtime feedback without turning the product into infrastructure theater.

PostgreSQL / Redis / MongoDB / ChromaDB / FAISS / Next.js / React / WebRTC

> 006 / Writing

Field notes from production systems.

Architecture essays for production AI, realtime systems, fintech infrastructure, and the operating choices that decide whether software survives real usage.

View all posts ->

2026-06-01

The production AI systems audit

// A field note on choosing the first AI workflow worth building before a team burns cycles on demos, tools, and unclear ownership.

6 min read

2026-05-18

Scaling WebRTC architecture for 1,000+ concurrent VTubers

// How a custom SFU topology, simulcast, and explicit latency ownership helped avatar-driven rooms scale without turning presence into lag.

5 min read

2026-05-05

Building startup software teams from zero to one

// What technical leaders need to make explicit before hiring turns a small founding team into a fragile delivery organization.

7 min read