Decision infrastructure for high-stakes systems

Decision Process
as a Service (DPaaS)

Every intelligent system does what a person does under uncertainty — observe, reflect, analyze, decide, act. LuckMa turns that loop into auditable, explainable infrastructure. AI and ML are enablers inside it, not the product.

The LuckMa decision process — observe, ask, analyze, decide, act, in a continuous loop

The runtime · interactive

Explore the loop.

Click a stage to see what it does. Trading is our first adapter — the loop is domain-agnostic.

01 · OBSERVE

Perceive the world, cleanly.

Ingest and normalize any signal — prices, events, records — into a single timestamped view, with bad-input rejection built in. Garbage never reaches the decision.

auto-cycling — click any stage to focus

The reach

Same core. Different adapters.

A new industry is new Observe (their data) and Act (their systems) around the identical loop.

Stage	Trading built	Credit	Fraud	Clinical ops
Observe	market data	applicant + bureau	transaction stream	vitals + history
Reflect	tradable moment?	complete file?	anomalous?	triage moment?
Analyze	P(win), E[R]	default probability	fraud score	deterioration risk
Decide	E[R] > hurdle	return clears risk	cost-weighted risk	risk × cost of miss
Act	place / hold	issue / decline	block / allow	escalate / monitor

The pitch

The decision is the product.
AI is just the enabler.

Companies already have models. What they lack is the disciplined process to deploy one safely against their incumbent rules, prove it earns trust, and audit every call. That gap is LuckMa.

01 · THE PROBLEM

High-stakes decisions can't be trusted to a black box.

In finance, credit, fraud, and health, a wrong automated decision has real cost — and a regulator asking why. "Our AI decided" is not an answer. Teams either over-trust models or never ship them.

02 · THE THESIS

Abstract the decision itself.

Observe → Reflect → Analyze → Decide → Act — the loop every good decision-maker runs under uncertainty, delivered as auditable, explainable infrastructure. The enabler underneath (rules, ML, LLMs) is swappable; the loop is invariant.

CALIBRATED

Confidence you can trust — scores that actually rank outcomes.

EXPLAINABLE

Every decision carries its reason. Audit by construction.

REVERSIBLE

Safety gates and a kill-switch. Nothing acts on faith.

03 · HOW AI EARNS TRUST

It's never trusted on faith — it earns its way in.

Each rung is separately sellable. Start rules-only today; reach autonomy only when the model has proven it beats the rule. A breach reverts instantly.

04 · THE VALUE

Value re-rates at milestones — not at more code.

05 · THE STRUCTURE

One lighthouse. One platform.

LUCKMA.IO — THE PROOF

Autonomous trading, our own capital

A live, ML decision system in the hardest arena there is: adversarial, real-time, real money. It proves the platform works.

LUCKMA.AI — THE PLATFORM

decisioning, for anyone

The same discipline sold as infrastructure to any team making high-stakes decisions — finance, credit, fraud, health.

06 · MACHINE LEARNING AS A SERVICE

A model for every stage of the decision.

Not one monolithic black box — a model per stage, trained in a cascade. Each model learns from the raw evidence and from the model before it, then passes its read downstream. The result is a calibrated, end-to-end decision you can inspect at every step.

OBSERVE

encodes the raw tape into a compact market-state read

→

REFLECT

screens whether this moment is even worth a decision

→

ANALYZE

reads structure and trend from the sequence

→

DECIDE

scores the setup — calibrated probability and expected value

→

ACT

chooses the action, or holds — with full provenance

Served as a local API — one model version per stamp, so every live decision is reproducible. New models ship in shadow first: they run alongside the incumbent, are journaled, and only take over once they beat it out-of-sample on both training and unseen data. A drift breach reverts instantly.

Founder · Principal Software Architect

Jae Lim (Ja Hon Sim)

A decade architecting large-scale systems at Google, Salesforce, eBay, and Yahoo — now applying that rigor to the hardest problem in AI: making high-stakes decisions you can trust.

GoogleSalesforceeBayYahoo

AWS Machine Learning AWS Big Data PCAP Python SCJP Java

Built to decide under uncertainty.

"The decision loop isn't a metaphor for me — it's how I've navigated my whole life: act well when the information is incomplete and the cost of error is real."

Jae came to the U.S. from South Korea at fifteen without a word of English. Through teachers, family, internships, and relentless self-teaching, he built a path through computer science, software architecture, and AI — shaped by a lifelong instinct for reading a situation with insufficient information and acting anyway. As a principal engineer in Silicon Valley he designed distributed systems at global scale; that same discipline now underpins LuckMa's decision platform.

Competitive Warcraft III — #1-RANKED WARCRAFT III · AMD-SPONSORED — READING THE OPPONENT IN REAL TIME

Decisioning as a discipline, not a slogan

A #1-ranked pro gamer's edge is the same as a trading system's: perceive fast, weigh the evidence, commit under pressure, adapt. At seven he beat Japanese RPGs he couldn't read — purely by inference and trial-and-error. That intuition for acting well without complete information is the thread from his biography to the trading engine to luckma.ai.

Contact

Let's talk decisions.

Interested in the platform, a design-partner pilot, or the trading system? Send a note — it goes straight to Jae.

@

Email

jae.lim@luckma.io

◈

Company

LuckMa LLC · Irving, Texas

↗

Focus

decision intelligence

Prefer email? Reach Jae directly at jae.lim@luckma.io. Responses within a business day.

Machine learning

The intelligence
behind the decision.

Deep neural networks trained on time-series big data — learning not just to predict, but to decide. Supervised where we have labels, self-supervised and reinforcement-learned where the market is the only teacher.

"When I saw move 37, I was proven wrong. AlphaGo is indeed capable of making creative decisions — an artistic move that resembles the beauty of Go."

— Lee Sedol, Go World Champion (9 Dan). Creative decisioning under uncertainty is not science fiction. It's engineering.

SEQUENCE MODELS

RNNs on the tape

We structure data nodes sequentially along a temporal series — recurrent networks that read the market the way it actually unfolds.

REINFORCEMENT

Learn to act

Agents take actions in an environment to maximize cumulative reward — the natural frame for a system that decides, not just forecasts.

SUPERVISED + SELF

Every signal a teacher

Where outcomes exist, we learn from them directly; where they don't, the engine's own structure becomes a free, noise-tolerant label.

Tuning hundreds of models — without overfitting.

Managing many models on big data doesn't scale by hand. We automate hyperparameter search and apply orthogonal optimization — tune one axis at a time (fit training, then validation, then holdout, then the real world), so each knob has a clear job. Naïve early-stopping isn't orthogonal; it trades training fit for validation fit at once, and hides the cause.

Recurrent neural network over a temporal sequence — RECURRENT NETWORK — DATA NODES ALONG A TEMPORAL SEQUENCE

Overfitting vs generalization — THE CENTRAL TENSION — FIT THE DATA, NOT THE NOISE

Cloud architecture

Three pipelines,
one cloud.

Data engineering, model training, and inference — built end-to-end on AWS. Data streams in from hundreds of sources, is transformed at scale, trains GPU models, and serves live inference in milliseconds.

LuckMa AWS data-engineering, training, and inference architecture

The assembly line.

INGEST

Hundreds of sources

An asynchronous ingestion engine receives and processes data in parallel — nothing waits in line.

APACHE BEAM

Parallel batch

Batch processing in a multi-threaded context so training data is prepared at scale, not one row at a time.

APACHE SPARK

Transform at scale

Master/worker clusters and a PySpark context transform feature vectors — over 500% faster than the naïve path in our benchmarks.

DASK

Vectorized extraction

Feature extraction that leans on vectorized operations — up to 100× over a Pandas apply in our tests.

SAGEMAKER · GPU

Train fast

GPU-enabled EC2 (p/g) with Nvidia GPUs and Pipe-mode streaming for robust, high-throughput training.

TF SERVING · LAMBDA

Serve in ms

TensorFlow Serving GPU-compiled with Bazel, invoked through AWS Lambda over gRPC/REST for low-latency inference.

Data security

Encrypted before it leaves.

Before training data ever hits the wire, we apply proprietary encryption — so sensitive inputs are protected in transit and at rest, without giving up the ability to compute on them.

Homomorphic encryption — COMPUTE ON ENCRYPTED DATA — SECURITY WITHOUT A BLIND SPOT

Cloud infrastructure

GPU-accelerated.
Infrastructure as code.

GPU-enabled AWS EC2 with Nvidia GPUs and deep-learning AMIs — engineered for scalability, 99.9%+ uptime, and security, with latency kept to a minimum.

GPU compute

Built for parallel math.

The core operations of deep learning parallelize beautifully — and GPUs have far more cores than CPUs. On our benchmarks GPU training runs 4–5× faster than CPU. GPUs are accessible directly inside EC2 p/g instances via AWS deep-learning AMIs.

Nvidia GPU — NVIDIA GPUs — MANY CORES, MASSIVELY PARALLEL

Docker with AWS SageMaker — DOCKER + SAGEMAKER — INFRASTRUCTURE AS CODE

Docker · SageMaker · IaC

Reproducible by design.

Every environment is code. SageMaker pulls a Docker image from AWS ECR, trains on EC2, and uploads model artifacts to S3 — the whole pipeline versioned and repeatable, never a hand-configured server.

CUDA · nvidia-docker

GPUs, straight through the container.

Nvidia CUDA provides the GPU-accelerated libraries deep-learning frameworks depend on — and TensorFlow pins a specific CUDA version, so the stack has to match exactly. Our Dockerized inference pipeline integrates nvidia-docker for GPU pass-through of the CUDA driver between host and container, so we train and serve on GPU inside containers that stay agnostic to the host OS.

99.9%+ UPTIME

Redundant, scalable EC2 fleets.

LOW LATENCY

Inference co-located with the GPU.

SECURE

Encrypted data, least-privilege access.

Decision Processas a Service (DPaaS)