Cognitive dispatch copilot
LLM-orchestrated exception handling with tool-calling into TMS, slot booking, and driver comms— guardrailed, auditable, multilingual.
Neural ops · AI-native control mesh
- RAG control towers
- Digital twin mesh
- Agentic dispatch
- Federated sensing
- Real-time vector memory
Manifest · fully digital freight
The next decade is not “more TMS screens”—it is a sovereign, self-hearing corridor
Tomorrow’s transport is a closed-loop cyber-physical system: every trailer, swap-body, customs broker, and energy node emits signed telemetry. ALARA treats that stream as a single spatiotemporal tensor—compressed, replicated, and reasoned over with models that understand delay propagation the way compilers understand register pressure.
Physical internet semantics (modular load units, automated hand-off cells, predictive slotting) merge with digital bills of lading, CEP-5G network slices, and post-quantum handshakes between carriers. We design for a world where convoys negotiate priority through verifiable auctions, where carbon is a first-class constraint, and where a missed handshake triggers not a phone tree but a self-healing DAG of compensating actions.
ALARA’s north star: zero idle reasoning—every millisecond of compute either predicts a disruption before it crystallizes, or compresses operator intent into an auditable action. This is the operating system layer for logistics that finally matches the complexity of the physical world it moves.
Neural capability layer
Synthetic operations · human-in-the-loop when it matters
ALARA architects retrieval-augmented operations that ground every decision in live TMS, telematics, and customs telemetry. We orchestrate multimodal embeddings for route-risk, deploy probabilistic digital twins for what-if capacity, and tune closed-loop reinforcement policies on your KPIs—latency, carbon, SLA, margin—without ripping out legacy stacks.
Transport OS · 2035-grade
Corridors, convoys, and compliance as one executable specification
Autonomous convoy contracts
Platooning and hub-to-hub automation expressed as formal convoy contracts: speed profiles, headway laws, and liability hand-offs compiled to on-board policy bundles. ALARA simulates entire weeks of mixed autonomy before a single wheel turns on a live motorway.
Intermodal tensor yards
Rail–road–shortsea yards modeled as 3D slot tensors with crane choreography, thermal drift on reefer strings, and tidal window coupling. Optimization runs continuously under stochastic dwell-time distributions imported from historical TEU shock libraries.
Border-grade reasoning
HS-code and sanctions logic interwoven with explanation graphs for customs: every clearance path carries proofs consumable by auditors and partner APIs. When regimes shift, the graph rewrites faster than PDF playbooks ever could.
Energy-aware routing
Fleet routes co-optimized for kWh / tonne-km, grid congestion, and hydrogen refill cadence—merged with live electricity futures where available. Carbon becomes a Lagrange multiplier, not a footnote slide.
Resilience mesh
When a port silo fails, ALARA spins parallel universes in twin space: thousands of reroute hypotheses ranked by Pareto fronts of cost, time, and emissions—surfaced to humans as a tight menu of defensible decisions.
Last-mile cognition
Micro-fulfilment and lockered cities represented as hyperlocal graphs with pedestrian wind, curb robots, and crowd-sourced hazard tags. ETA distributions shrink with every block-level observation ingested at the edge.
IT core · sovereign stack
Runtime, data plane, and control plane engineered as one artifact
- runtime WASM + eBPF sandboxes for tenant agents; hot-swappable inference graphs without dropping QUIC sessions.
- data_plane columnar event lake + immutable append-only ledger slice for compliance exports; CRDT-backed collaborative dispatch boards.
- control_plane GitOps for policy bundles; canary routing of model weights; chaos drills scheduled like cron for supply shocks.
- observability exemplar-based tracing with OpenTelemetry lineage into billing and carbon attribution—one trace ID from sensor to settlement.
- security SPIFFE/SPIRE identities per microservice; OPA/Rego policy gates on every tool-call an LLM attempts; confidential VMs for partner co-processing.
- delivery progressive rollout by corridor percentage; shadow traffic on twin stacks before promotion to production lanes.
Capability matrix
Primitives composable into operator-grade systems
Predictive ETA & demand sensing
Chronos-style sequence models + graph neural nets on corridor topology for sub-15 min ETA bands and load-level demand shock alerts.
Digital twin & simulation mesh
High-fidelity synthetic yards and hubs for Monte Carlo stress tests, peak-season choke forecasting, and autonomous reroute rehearsal.
Edge inference & telemetry fusion
On-device vision + CAN fusion, sub-second anomaly scoring, and offline-first queues synced through CRDT-friendly meshes.
Agentic workflow fabric
Multi-agent planners for dock scheduling, customs pre-clear, and claims—policy-as-code with human escalation paths and full trace graphs.
Zero-trust transport mesh
mTLS service identity, OAuth2.1 / FAPI-grade broker patterns, and confidential compute lanes for partner data without centralizing PII.
Geospatial neural routing
Rasterized hazard tensors (weather, strikes, roadworks) convolved with fleet state for dynamic corridor scoring beyond static distance matrices.
Settlement & freight finance
Event-sourced accruals tied to proof-of-delivery cryptography; smart-contract templates for demurrage and cross-border escrow simulation.
Reefer & hazmat digital threads
Continuous validation chains for temperature, seal integrity, and ADR class drift with automated regulatory packets per jurisdiction hop.
Multilingual ops corpus
Fine-tuned retrieval over SOPs, incident logs, and driver comms—jurisdiction-aware lexicons for CEE, Nordics, and UK English variants.
Scenario blast furnace
Parameter sweeps across fuel price shocks, canal closures, and strike calendars—exported as executive survival surfaces (cost vs. service vs. risk).
Corridor federation gateway
Versioned partner APIs with contract testing in CI, traffic shadowing, and automated deprecation windows so integrations never rot silently.
Frontier lab
Where transport physics meets machine reasoning
We prototype neural operators on irregular road graphs, diffusion priors for yard congestion, and self-supervised telematics codecs that beat generic gzip on mixed float/categorical streams. Publication-grade rigor is optional; operator-grade repeatability is not.
Roadmap tracks include quantized fleet federated learning with per-tenant noise budgets, causal impact models for policy changes in customs regimes, and neuro-symbolic slotting that respects hard safety constraints while still exploring combinatorial search spaces.
Founder · CEO
ALARA is led by Alessia Albu — builder, researcher, and ICHB academic trajectory
Alessia Albu is the Founder & Chief Executive Officer of ALARA, shaping this venture’s product vision, research posture, and ethical guardrails for AI in logistics.
Born in 2007, she studied in Bârlad at “George Tutoveanu” school until grade 6, then entered The International Computer High School of Bucharest (ICHB) from the 2020–2021 school year onward—documented publicly on her official site, alessia.pro. While advancing through ICHB she built ALARA from the ground up as a digital logistics and IT initiative: systems thinking, interface craft, and narrative clarity converging into what you see here.
“My dream is to study archaeology at UCL (University College London).”
Her public record includes the World Scholar’s Cup London 2023 (Global Round at UCL), where she earned three gold and two silver medals; the Tournament of Champions at Yale (2023) with a team gold and an individual silver; exhibitions such as THE HOLY ART Gallery, London (July 2023); and international performances with IFLC (e.g. Houston, 2024) and other programmes she lists on alessia.pro.
For educators: Alessia’s path is one of measurable international milestones reached while carrying a full ICHB workload—evidence of discipline transferable to research labs and industry product teams alike.
Pentru profesori și mentori
Alessia Albu este elevă la Liceul Internațional de Informatică din București(ICHB), cu un parcurs public documentat pe site-ul personal alessia.pro, inclusiv competiții precum World Scholar’s Cup London și Tournament of Champions, Yale, U.S.A.. Este Founder & C.E.O. ALARA, proiect construit de la zero în paralel cu liceul, la intersecția dintre transport digitalizat, inginerie software și comunicare de precizie. Pentru fotografii și cronologie detaliată, sursa oficială rămâne alessia.pro
Press & public milestones
Third-party and institutional references (not exhaustive)
- SparkNews.ro — ICHB students, Africa volunteering (June 2023)
- alessia.pro — timeline, medals, exhibitions, IFLC
- ICHB Facebook posts — linked from Alessia’s timeline on alessia.pro (World Scholar’s Cup Bucharest 2023, etc.)
- Honorary Consulate of Romania to Connecticut — Facebook photo reference (Yale TOC 2023), linked from alessia.pro
Secure uplink
Encrypted handoff · no public mail surface