Named systems
we've
shipped.

Operators of on-site teams were running attendance and compliance on manual timesheets, screenshots, and manager memory — producing payroll disputes, late exception discovery, and no live picture of who was actually on location. Onsight needed a workforce intelligence platform that could turn every attendance event into verifiable, policy-bound, audit-ready operational truth — combining geofencing, trusted-device controls, offline-tolerant mobile capture, and an AI command layer that could explain exceptions in plain language.

A four-surface platform — admin command center, native mobile app, exception action-queue, and the OpsIQ AI intelligence layer — built end-to-end. Geofence-verified check-ins and check-outs against approved worksites, biometric and trusted-device controls on eligible plans, live workforce visibility bound by working-hour policy, and offline-resilient mobile sync. OpsIQ turns attendance events, trust signals, and audit logs into AI-explained briefs managers can act on before payroll closes.

• Geofence-verified attendance with trust-scored check-ins
• Live workforce visibility during working hours
• Exception intelligence — missed checkouts, breaches, anomalies
• Offline-resilient mobile capture with queued sync
• OpsIQ AI explanations for exceptions and audit activity
• Biometric & device-integrity controls (Enterprise tier)
• Payroll-ready attendance exports

Kreda's core SaaS platform had grown around a large monolithic application backed by RethinkDB. As adoption increased, high-traffic campaigns placed pressure on the platform, while RethinkDB's inactive ecosystem created long-term operational risk, hiring constraints, and maintenance concerns. The migration could not be treated as a simple export/import: years of production data, hidden business logic, document-model queries, implicit relationships, and zero tolerance for data loss made the project mission-critical.

Hexcore modernized Kreda's persistence layer by migrating production data from RethinkDB to Amazon RDS MySQL. The engagement covered production backup to Amazon S3, codebase reverse engineering, relational schema design, backend refactoring, deterministic Python migration tooling, multi-stage validation, and repeatable migration runs. The internal migration framework was generalized into Rethinkport, an open-source tool published to PyPI for teams moving from RethinkDB into relational databases.

• Complete RethinkDB production export secured in Amazon S3 before transformation
• Application code review to recover implicit relationships, validation rules, and query behavior
• Relational schema redesign with primary keys, foreign keys, composite indexes, unique constraints, and views
• Deterministic Python migration pipeline for repeatable test and production runs
• Validation stages covering record counts, data types, required fields, duplicates, and referential integrity
• Backend refactoring from document-model assumptions to a maintainable relational foundation
• Open-source migration framework generalized as Rethinkport and published to PyPI

Hospitals, fulfillment centers, and retail floors were running rosters on spreadsheets, paper, and HRIS calendars that were never designed to think about skill mixes, credential expiry, multi-jurisdiction labor law, and last-minute call-outs in time. Shift leads were burning 14 hours a week rebuilding the schedule, one in five shifts was filled by unbudgeted overtime or 2.4× agency cost, and frontline turnover linked to schedule fairness was running at 38%. The teams keeping the real economy running needed a workforce platform that could absorb human reality — not pretend it away.

An AI-native workforce platform that turns skill mixes, labor law, and last-minute chaos into a roster operators can actually publish. A CP-SAT solver enforces 142 hard and soft constraints — credentials, ratio rules, union clauses, fairness, commute, sleep cycle — and produces a fully explainable roster in 1.8 seconds. Three purpose-built models (demand forecasting, no-show prediction, burnout & fairness) trained on each customer's historical shift data continuously evaluate against ground truth. Customers go live in 14 days, publish their first AI-generated roster inside week two, and cut roster-build time by 63% while saving $1.2M of overtime per 1,000 FTE annually.

• CP-SAT optimization engine — 142 constraints, multi-objective, 1.8s solve
• Real-time replan: re-solves the affected slice in under 2 seconds on a call-out
• Demand-forecasting model — hour-by-hour by location, role, skill mix (MAE 4.6%)
• No-show prediction — calibrated per-employee, per-shift (AUC 0.87, recall 0.81)
• Burnout & fairness model — detects workload drift, cuts voluntary attrition −27%
• Explainable output — every assignment carries a reason; every rejection an alternative
• Industry packs: healthcare (acuity, ratios, credentials), warehouse (TEUs → labor), retail (Fair Workweek)
• Compliance engine: HIPAA, OSHA, Fair Workweek (5 US jurisdictions), CA meal/rest, EU Working Time
• Native iOS + Android worker app — shift view, swap, leave, clock-in, team chat
• Integrations: Workday, UKG, ADP, BambooHR, Kronos, Epic, Cerner, NetSuite, Snowflake, Slack, Teams
• SOC 2 Type II · HIPAA + BAA · ISO 27001 · GDPR-ready · SAML SSO + SCIM
• 14-day implementation playbook — ingest, constraint authoring, shadow run, go-live