SynectoAI · Agricultural lending · Credit risk
Connecting a headline to a loan, before the quarterly review.
The problem. A drought is declared across the Ohio corn belt. To act on it, a credit officer has to find which borrowers grow corn there, check where each delivers their harvest, verify whether a backup route exists, recalculate revenue against origination assumptions and re-run probability of default. Half a day per headline, across several systems — then repeat for every other headline that day. Mostly it doesn't get done: the headline goes in a folder, the loan waits for the quarterly review, and by then the borrower has missed a payment.
What we built. Watchtower ingests thousands of live news, weather, shipping, market and government feeds. Historian holds a time-aware knowledge graph of the value chain. Analyst extracts entities and events and traces multi-hop causal paths to a specific exposure. Briefer writes the alert, with a confidence score and a recommendation. Risk Output puts it in banking language — PD movement, basis widening, revenue impact.
Same headline, resolved end to end in under ten seconds — automated and always on. The credit officer decides; the system does the tracing.
Agentic decision systems · Causal intelligence · Knowledge graphs
WealthEngine.AI · Investment intelligence
Investment decisions with the analysis already done.
Deep reinforcement learning generates trading signals; neural networks and tree-based models work over derived features; a factor-scoring layer produces proprietary stock and fundamental ratings; a generative layer turns scores into readable portfolio insight. Value, momentum and small-cap dynamic portfolios ship volatility-targeted, hedged or leveraged. Agentic capabilities are in active development.
Patented AI/ML methods in financial markets — synthetic data generation for thin training data, and momentum trading via an ensemble of deep reinforcement learning algorithms. Peer-reviewed at HICSS-58, published by Springer.
Predictive & adaptive modelling · Generative AI
Healthtek · Healthcare · Revenue cycle
Turning paper EOBs into posted payments.
Explanation-of-Benefits documents still arrive on paper, in hundreds of payer-specific layouts, and every one has to be read and keyed by a person. Our patented models perform contextual extraction — learning from the client's own document set rather than needing a template — with deep learning locating regions of interest, key-value pairing preserving relationships, and validation gating the output as EDI 835 or any required format.
Up to 84% fewer manual postings. Turnaround inside 24 hours, a measurable fall in DSO, fewer errors than the manual process. HIPAA compliant, in production.
Document intelligence · Human-in-the-loop · Regulated data
Identity verification · KYC capability
Confirming the document is real and the person matches it.
A platform needed reliable verification of official identification documents at a volume manual review couldn't reach. We built computer-vision models that validate and authenticate government IDs — checking the document itself, not just reading text off it — and match the person presenting it against the identity it describes.
Verification suitable for regulatory-compliance purposes, at a speed manual review couldn't reach. The production evidence behind our KYC/AML capability.
Identity & entity intelligence · Computer vision
Entity resolution · Master data
One entity, forty names, one record.
The same entity appears under dozens of naming conventions across source systems. We built a layered stack — named-entity recognition, string and fuzzy matching, semantic similarity, an ontology layer and custom rules — where ambiguous cases surface as potential matches and route to a human reviewer instead of resolving silently.
Designed to a 99% precision and 99% negative-predictive-value specification — as reliable at saying "this is new" as at "this is a match". The engine behind KYC screening and claims deduplication.
Identity & entity intelligence · NLP · Human-in-the-loop
Document intake · Claims & underwriting pattern
From PDF to structured record, with a human in the middle.
High-volume inbound PDFs that must end up as structured records in an enterprise system. We built automated field extraction plus an annotation interface where a user verifies what the model pulled before anything is written through to SAP. The review step isn't a workaround — it's the design.
The same architecture we recommend for claims intake and underwriting submission ingestion, where fully autonomous extraction wouldn't pass a risk function.
Document intelligence · Human-in-the-loop · System integration
Group B — Applied AI across industries
Beyond our core verticals we've shipped AI in steel, industrial gas and manufacturing. The domains differ; the engineering — vision, compliance rule engines, evaluation, on-premise deployment — is the same stack we bring to regulated financial and clinical workflows.
Emissions compliance Solution blueprint
Teaching a camera to follow the regulator's own method.
Gas leakage from coke ovens is graded today by an inspector walking the battery, as often as someone can make the walk. We designed cameras at exactly the observation points the standard specifies; vision models detect emissions, attribute them to the right component and grade severity on the same four-point scale the inspector uses. A rule engine applies every exclusion in the standard before any metric is computed. Fully on-premise — no video leaves the site.
Reproduces the approved methodology without deviation, and shows its working. Every regulated reporting obligation has this shape — a bank reading this card is reading about regulatory reporting.
Vision-based compliance monitoring · AI assurance & governance · On-premise
a division of Data-Core Systems