AI Product Builder · Finance · Asia
At RM Investment Bank, I lead IT across Asia and remain hands-on as Lead Full Stack Developer. I turn high-friction finance workflows into reliable products used by real teams and paying customers.
Kuala Lumpur, Malaysia
By the numbers
$10M → $200M
AUM growth
Inter-Pacific Asset Management · ~1.5 years
$50K → $200K
Monthly net profit
Trader portal · within 1 year
~400%
Marginal profit lift
Salmi Niaga · systems and unit-economics discipline
15+
Production apps
Dashboards, signal bots, RAG agents, contest engines
1.4s → 80ms
p50 latency tuned
SprintBo bot /ranking · D1 + KV across 3 surfaces
< $5
Monthly infra cost
SprintBo · cron-driven, idempotent, zero ops headcount
What I lead and build
Live Mini App
01
Trading SaaS, investor analytics, market-data pipelines, and operator tools built for production decisions rather than demos.
Live Telegram roster
02
Production agents with retrieval, memory, multimodal inputs, and human intervention designed into the operating loop.
hydraxrail.com
03
Technology direction stays connected to the product, the team, and the production systems that have to keep working after launch.
Selected finance work
Two live flagships: a trader operating platform and a multi-persona RAG system grounded by bge-m3 retrieval and a knowledge graph. TradeClaw now runs in Research mode; stox.my remains part of the broader product record.
Live Mini App
Flagship · liveCase study
Solo · 1 year · replaced what would have been a 3-engineer + PM build
Live trader platform with AI signals, leaderboards, community automation, and a four-mode contest engine. React 19 + Hono, 15+ REST endpoints, Durable Objects rate limiting, and a tamper-evident credit-integrity audit trail. Lifted client trading-desk monthly net ~$50K → $200K within 1 year.
Specialist roster
Flagship · live
Solo · personas + retrieval + knowledge graph · live production system
Live multi-persona AI system combining bge-m3 RAG, a knowledge graph, query rewriting, tenant-aware routing, and durable follow-up memory. Specialist personas share grounded knowledge while keeping distinct roles and conversation styles.
tradeclaw.win
Research modeCase study
Solo · backend + frontend + vision-LLM · active research system
Research environment for multi-asset signal evaluation, vision-LLM chart parsing, backtesting, and track-record analysis. FastAPI + React with one consistent signal schema across Crypto, Forex, Commodities, Indices, and Equities.
stox.my
Investor analytics · 0-to-1Case study
Solo · founding engineer · 0-to-1 to paying users
Stock analytics & AI alpha-discovery SaaS for Malaysian markets. Real-time Bursa pricing, multi-factor screener, watchlists, AI-assisted alpha discovery. Edge-first, sub-100ms reads on hot queries. Founding engineer, 0-to-1 to paying users.
Supporting collaboration
I support product development with BizJaya Techlabs across FinQuest and FinNews. Devon Vejay is the primary contributor; my role is supporting rather than lead ownership.
More production work
Professional engineering proof without another media gallery.
Live · multi-agent runtime
Not a typical support bot. A dynamic multimodal runtime where multiple AI agents converse with each other in real time while humans observe and step in. Each agent holds its own persona, memory, and modality. Live Telegram bot — type /help to see the personas talk.
Agent orchestrationSolo · multi-agent runtime + per-persona memory · live to public users
Signals SaaS · live
Telegram signals across Crypto, Forex, Commodities, Indices, Stocks. On-demand chart breakdowns via vision-LLM parsing. Consistent schema across asset classes — one signal engine, five markets.
Five asset classesSolo · 5 asset classes on one signal engine and schema
Cost-disciplined SaaS
Bot /ranking p50 1.4s → 80ms via D1 source-of-truth + KV cache shared across bot + Mini App + landing. Operator pause/close/cancel with tamper-evident final_snapshots. Hosting under $5/month.
80 ms p50 readsSolo · bot + Mini App + landing on one worker · sub-$5/mo hosting
MAS-licensed onchain
White-label workspace for tokenised product issuance on MAS-licensed infra. Five role portals (Issuer, Distributor, Investor, Ops, Admin) with RBAC and audit-ready data model. Onchain via Canton smart contracts.
Five role portalsFull-stack engineer · 5 role portals · MAS-licensed infra
Open source · npm v3.9.2
Claude Code plugin + GitHub Action transcript linter. Skills, commands, hooks, agents, instinct packs, synthetic checks — structured self-improvement loops with eval-driven verification.
Eval-driven loopsSolo maintainer · npm package + GitHub Action · public eval harness
Reflection companion
Choose a verse. Slow down. Reflect with structure.
A calm, AI-assisted companion for personal Qur'an reflection, with guided entry points, thematic studies and saved reflections.
Product notes
Mosque marketplace
Find a mosque service without chasing separate contacts.
A Malay-first marketplace for mosque facilities, classes, community programs and vendors across Selangor and Kuala Lumpur.
Product notes
Source-grounded learning
Ask a question, then follow the source back to the book.
A bilingual study preview for Sheikh Dr. Sa'id Foudah's published works, organised around Ask, Learning Path and Library.
Product notes
Clinical communication practice
Practise the conversation before the exam room.
Voice-first ten-minute communication and ethics stations for MMed and PACES candidates, followed by rubric-based examiner feedback.
Product notes
Production evidence
Product captures and role-scoped portal evidence stay out of the way until you choose to inspect them.
6 production captures · 5 role portals
LiveTradeClaw
Verified track record · +129.66% total return · 42.1% win rate · 1,617 resolved signals
Recordedstox.my
Founding backer deck · live KLSE watchlist with MAYBANK, PBBANK, TENAGA, CIMB, IOI
LiveHydraxRail
Institutional marketing landing routing into five role-scoped portals · Canton-rail integration
LiveTradeClaw
Live Backtesting Engine · XAUUSD on H1 · equity curve, strategy comparison, real Binance + Stooq data
LiveTradeClaw
Signal Performance Leaderboard · 2,772 signals tracked across 26 pairs · per-pair hit rates
Livestox.my
Market Screener · daily-updated KLSE screening with structured fundamentals, technicals, and signal filters
HydraxRail · 5 portals
RecordedIssuer
Product feed, approvals, investor scope
RecordedDistributor
Subscription queue, approval ceremonies
RecordedInvestor
Available products, subscription state
LiveOps
Audit trail, escalations, SLA timers
RecordedAdmin
Tenant onboarding, IdP and SSO glue
Career line
2026 — now
Lead technology direction across Asia while staying hands-on with product architecture and delivery. The remit spans AI, software, digital products, team execution, and the production systems behind them.
2024 — 2026
Built a trader portal (apps, AI signal tools, leaderboards, community automation) that lifted average monthly net ~$50K → $200K within 1 year — replaced what would have been a ~3-engineer + PM build. 15+ production apps across investor dashboards, signal bots, RAG agents, weekly contest engines.
2023 — 2024
Founded an AI-first venture. Secured seed investment from Cradle and Nexea. Owned product, engineering, and GTM end-to-end across Python, TypeScript, React, PostgreSQL, AWS — first-engineer posture from day one.
2021 — 2023
Drove ~400% increase in marginal profit through better systems, automation, and execution discipline. Built Python analytics dashboards and reframed P&L from gross-revenue chasing to unit economics.
2020 — 2021
Contributed to AUM growth from ~$10M to ~$200M in ~1.5 years through data-driven research and screening. Built data pipelines, factor models, and analytics dashboards in Python and SQL.
Best Shariah Equity Malaysia 2020 — Lipper / Refinitiv
2016 — 2020
Fund accounting, treasury operations, financial modelling. Excel/VBA automation that compressed reporting cycles — first hands-on exposure to the data shape behind investor analytics.
Current stack
LinkedIn-confirmed core skills plus current project-proven tools, organized by the systems they support.
Agents, retrieval, evaluation, and production inference.
Web and mobile surfaces, services, and product architecture.
Market data, APIs, transactions, and analytical workloads.
Edge-first delivery, cloud operations, and repeatable releases.
Contact
I read every message. Responses typically same-day on weekdays. If it is not a fit, I say so directly and point you to someone who is.
For agency or NDA briefs, send a one-paragraph summary first.
2026 — now
2024 — 2026
2023 — 2024
2021 — 2023
2020 — 2021
2016 — 2020
