Sohel Sayed
Portrait of Sohel Taher Sayed

Sohel Taher Sayed

Product leader who builds. AI agent systems, blockchain products, trading automation and full platforms.

12 years shipping for HSBC, MetLife and AXA through EY, IBM and Capgemini. Now in Dubai, leading product for two live blockchain platforms.

An engineering team made of agents.

I designed and run a system where AI agents take an idea from spec to shipped feature. I make two decisions. The agents do the rest, and prove it.

Feature pipeline, illustrativerunning
How a feature moves through the agent system A spec is planned and approved by me, built by three parallel agent lanes, checked by a council of reviewer agents, tested, passed through a safety gate and shipped. Findings loop back to design. A shared memory feeds every stage. finding, back to design Ideaa spec PlanI approve Build agent Build agent Build agent Review council Testproves it Gate unsure? it stops Shipwith evidence Shared long-term memory: every decision, lesson and correction
Ideaa spec PlanI approve Build Build Build Review council Testproves it Gate unsure? it stops back to design Shipwith evidence Shared long-term memory

Parallel agents

Work splits across agents that build and test side by side.

A review council

Independent reviewers, security included, must agree before anything moves.

Memory that compounds

Lessons persist for months. It does not repeat a mistake it has learned from.

Rules, not vibes

When to stop and ask is a written rule, never the model judging itself.

Works overnight

Runs unattended inside limits I set. Anything risky is refused, not attempted.

Grades itself

Audits itself constantly, and is allowed to give itself a failing grade.

~30%faster delivery
9severe defects stopped
before release
2human decisions
per feature
$0per-call AI bills

Genie

An always-on agent that acts as tech lead. It drives work to done and only interrupts me at a real decision.

Aria

My personal assistant on Telegram. It plans, triages and briefs me, and tracks whether its suggestions are actually useful.

Blockchain

I lead product for two live platforms at LMNTO, Dubai, since 2025.

DePIN rewards platform

LMNTO, 2025 to now

Live

Households run a small device that shares spare internet capacity for token rewards. I own the roadmap, wallet, KYC and compliance.

  • 19,000+sign-ups
  • 5,000+active users, 84 countries
  • $1.2M+deposits
  • +40%conversion from my Fast Track programme

DAO governance platform

LMNTO, 2025 to now

Live

Token holders propose and vote, and outcomes execute on chain. I wrote the V2 governance contract myself.

  • 24+proposals
  • ~800lines of Solidity I authored
  • 0critical or high security findings

Token-reward simulator

LMNTO

My Python model caught a reward plan paying 43% against a 34% cap, before launch. Stakeholders still price every change with it.

Trading automation

Tools that move or watch money, built to be measured.

Copy-trading engine

Personal, 2026

Live

Mirrors trades from signal-provider strategies into a separate MT5 account in real time. Each strategy runs on its own rate budget, so one busy strategy never slows the rest.

  • <1strade detection
  • 4 to 12strategies in parallel

Python, MetaTrader 5, Telegram alerts

Treasury monitor

LMNTO, open source

Live

Twice-daily wallet reports across Polygon, Arbitrum and TRON. Read-only, so it never holds a key.

Executive dashboards

LMNTO, open source

Live

Analytics from inside a VPN, delivered safely to 11 executives, the CEO included.

Full platforms

Specified, built and tested end to end.

Housing-society platform

Personal, 2026

Live

Everything a paid society manager does: billing, accounts, visitors, complaints, meetings and dues recovery. Resident protections are enforced in the database, not on the screen.

  • 99web routes
  • 15rules enforced in the database
  • 4languages

VoiceEarn

Personal, 2026, open source

Live

People read published books aloud and get paid per approved page, in USDC on Polygon. Includes a guided reading room and a full admin review console.

React, Supabase, serverless Node.js, Polygon

Process improvement

Six changes, each with its evidence.

~85%

faster policy issuance at MetLife UAE, from days to instant.

82

design gaps caught in a build where all 504 tests passed. Now gated.

700,000

records migrated off a mainframe with zero data loss.

40+

process gaps found before a line of the new platform was built.

9 pts

payout overrun caught in simulation, before launch.

2

manual finance and leadership reports now run themselves.

Experience

12 years, four companies.

LMNTO Performance

2025 to now, Dubai

Lead Product Manager

Leads product for two live blockchain platforms. Brought AI into how the company builds.

Ernst & Young

2022 to 2025, Mumbai

Managing Consultant, Product Lead

Merged AXA Singapore's platforms into HSBC Life after a ~US$529M acquisition: 1,000+ agents, 200,000+ policyholders.

IBM

2017 to 2022, Mumbai

Business Transformation Consultant

Led MetLife UAE's digital platform across 36 channels. Cut policy issuance ~85%.

Capgemini

2014 to 2017, Mumbai

Consultant

Moved 700,000 policy records off a mainframe with zero data loss.

Skills

Product

Strategy, roadmaps, discovery, prioritisation, KPIs, go-to-market

AI

Agentic systems, multi-agent review, evaluation, retrieval, model routing

Blockchain

DAO governance, tokenomics, DePIN, wallets, Solidity, Hardhat

Insurance

New business, underwriting, claims, policy admin, straight-through processing

Transformation

Post-merger consolidation, legacy migration, Agile, SAFe

Building

Python, SQL, JavaScript, Node.js, React, Playwright, Git

Education

MBA

Symbiosis, Pune

MCA

Sikkim Manipal

B.Sc. IT

University of Mumbai

LOMA ALMI

Life and health insurance

Say hello.

Always up for a conversation about product, AI or blockchain.

sohel@sohelsayed.com

LinkedInGitHub