Historical task source. The Eve task board holds current task state.
Phase 130: Kalika Autonomous Experimentation — DOE, Robotic Synthesis, Instrument Control, Safety, Sample Operations & Closed-Loop Scientific Discovery#
Kalika's current materials research platform can ingest experiments, align them with simulations, and recommend synthesis routes, but the platform is not yet a true operating system for frontier experimental science. Leading groups do not just upload spectra and microscopy after the fact: they plan campaigns, allocate scarce instrument time, author executable protocols, track sample lineage, enforce SOP and hazard constraints, drive instruments and automation, and close the loop between predictions and measurements in near real time. This phase makes Kalika experimentally sovereign. It extends Phase 128's lab-data and collaboration model with full design-of-experiments, autonomous lab planning, robotic execution, instrument orchestration, sample inventory, facility operations, and audit-ready experimental governance.
130.1 Scaffolding & Core Types (@kalika/experiment-orchestration)#
130.1.1 Project Setup#
- 130.1.1.1 Create
libs/kalika/experiment-orchestration/with standard Nx configuration, TypeScript orchestration layer, Rust services for queue and protocol scheduling, and Python adapters for lab automation ecosystems - 130.1.1.2 Create sub-libraries:
@kalika/doe,@kalika/protocols,@kalika/instrument-control,@kalika/sample-ops,@kalika/lab-safety, and@kalika/closed-loop-experiments - 130.1.1.3 Link the experimental stack to Phase 128 notebooks, collaboration, ELN/LIMS connectors, discovery agents, and knowledge graph
130.1.2 Experimental Operating Model#
- 130.1.2.1 Define
ExperimentCampaigntype: scientific objective, design space, hypotheses, success criteria, budget, facility constraints, personnel, and status across planning/execution/analysis/publication - 130.1.2.2 Define
ExecutableProtocoltype: ordered steps, required inputs, labware, reagents, equipment, tolerances, branching logic, machine/human actions, pause/resume points, and QC gates - 130.1.2.3 Define
SampleLineageGraphtype: precursor lots, process steps, derived aliquots, mounting/packaging state, destruction/archive events, and links to all associated measurements and simulations - 130.1.2.4 Define
InstrumentSessiontype: reservation, operator, instrument configuration, calibration state, environmental controls, acquisition parameters, raw data locations, and exception log - 130.1.2.5 Define
SafetyEnvelopetype: chemical hazards, process hazards, PPE requirements, waste stream, interlocks, approval rules, and allowed automation boundaries
130.2 Experimental Design & Active Campaign Planning (@kalika/doe)#
130.2.1 DOE Foundations#
- 130.2.1.1 Implement classical DOE generation: full factorial, fractional factorial, Plackett-Burman, Taguchi, Latin hypercube, D-optimal, and response-surface designs
- 130.2.1.2 Implement constrained DOE: forbidden combinations, compositional simplex constraints, process-window limits, equipment throughput, and precursor inventory limits
- 130.2.1.3 Implement measurement plan synthesis: for each candidate experiment, select required characterization sequence, replicate count, calibration frequency, and stopping criteria
130.2.2 Bayesian & Sequential Experimentation#
- 130.2.2.1 Implement Bayesian optimization for experiments: expected improvement, knowledge gradient, entropy search, constrained BO, and multi-objective acquisition functions
- 130.2.2.2 Implement active-learning campaign planner: use uncertainty from ML models, process surrogates, and prior experiments to select the next most informative experiments under time/cost constraints
- 130.2.2.3 Implement adaptive replicate allocation: decide when to spend budget on replication vs. exploration based on noise estimates, heteroscedastic uncertainty, and current decision boundary
- 130.2.2.4 Implement experiment portfolio scheduling: mix cheap/high-throughput screens with expensive/high-fidelity validation runs to maximize information gain per calendar day
130.2.3 Planning Outputs#
- 130.2.3.1 Generate campaign-ready plate maps, batch sheets, instrument queue plans, sample IDs, and material requisition lists from DOE output
- 130.2.3.2 Generate experiment rationale report: why each run was chosen, expected information gain, associated risks, dependencies, and fallback options
130.3 Protocol Authoring & Robotic Execution (@kalika/protocols)#
130.3.1 Protocol DSL#
- 130.3.1.1 Implement protocol DSL for synthesis, sample handling, annealing, mixing, deposition, curing, transfer, and characterization workflows with strongly typed step schemas
- 130.3.1.2 Implement protocol compiler: translate high-level protocols to target backends such as liquid handlers, furnace controllers, robotic arms, deposition tools, and operator checklists
- 130.3.1.3 Implement protocol simulation and dry-run mode: validate timing, resource collisions, unsafe transitions, unavailable consumables, and impossible labware movements before execution
130.3.2 Execution & Recovery#
- 130.3.2.1 Implement execution engine for protocols with event-sourced logs, pause/resume, retries, and checkpointing at safe interruption points
- 130.3.2.2 Implement mixed autonomy execution: machine-only steps, human-only steps, and collaborative handoff steps with signoff requirements and acknowledgment of critical actions
- 130.3.2.3 Implement exception recovery: failed dispense, precursor depletion, instrument timeout, out-of-tolerance temperature ramp, vacuum failure, or contamination event
- 130.3.2.4 Implement execution evidence capture: parameter traces, sensor streams, images, timestamps, and operator notes attached to each protocol step
130.3.3 Backend Connectors#
- 130.3.3.1 Implement liquid-handler protocol connectors: plate maps, dispense commands, dilution trees, and contamination-aware tip management
- 130.3.3.2 Implement deposition and thermal-process connectors: spin coating, sputtering, evaporation, CVD/ALD, furnaces, hotplates, gloveboxes, and pressure reactors
- 130.3.3.3 Implement robotic motion and sample-transfer connectors: tray movement, barcode/RFID confirmation, pick/place operations, and guarded transfer into controlled environments
130.4 Sample Operations, Inventory & Chain of Custody (@kalika/sample-ops)#
130.4.1 Inventory & Materials Handling#
- 130.4.1.1 Implement reagent and consumable inventory: lot tracking, expiry, purity, storage conditions, reorder thresholds, and compatibility with hazards and facility rules
- 130.4.1.2 Implement sample registry: automatically generate globally unique sample IDs, barcodes/QR/RFID tags, physical location, packaging state, and archival/disposal status
- 130.4.1.3 Implement plate, wafer, pellet, coupon, foil, powder, and thin-film sample container abstractions with geometry-aware position maps
130.4.2 Lineage & Scheduling#
- 130.4.2.1 Implement chain-of-custody event model: who created, moved, processed, measured, split, combined, or destroyed a sample, with digital signatures where required
- 130.4.2.2 Implement sample batching and queueing: group compatible samples for characterization, annealing, or beamtime to minimize setup overhead and maximize throughput
- 130.4.2.3 Implement contamination and incompatibility rules: prevent invalid co-processing of air-sensitive, radioactive, biohazardous, or mutually reactive materials
130.5 Instrument, Facility & Beamline Orchestration (@kalika/instrument-control)#
130.5.1 Device Abstraction Layer#
- 130.5.1.1 Implement
InstrumentDriverabstraction: connection lifecycle, capability discovery, command schema, state polling, health status, and safe shutdown - 130.5.1.2 Implement standard acquisition adapters for XRD, Raman, PL, UV-Vis, XPS, SEM/TEM/STEM, AFM, electrochemistry, thermal analysis, and diffraction/microscopy mapping workflows
- 130.5.1.3 Implement support for beamline and facility sessions: reservation metadata, visit proposal IDs, sample manifests, transfer to facility data stores, and post-run archival synchronization
130.5.2 Calibration, QC & Acquisition#
- 130.5.2.1 Implement calibration scheduling and enforcement: reference standards, drift checks, lamp/current normalization, alignment validation, and blocked execution when calibration is stale
- 130.5.2.2 Implement acquisition templates: reusable parameter presets for common experimental runs with validation against instrument capability and sample hazard constraints
- 130.5.2.3 Implement in-run QC: signal-to-noise thresholds, peak/reference checks, focus/alignment checks, and automatic reacquisition when quality is below threshold
- 130.5.2.4 Implement streaming acquisition hooks: push partial datasets into Phase 128 dashboards and closed-loop campaign planners before the full run is complete
130.6 Safety, SOPs, Compliance & Human Governance (@kalika/lab-safety)#
130.6.1 Safety Rules Engine#
- 130.6.1.1 Implement SOP and hazard knowledge base: process restrictions, chemical incompatibilities, exposure limits, PPE, waste handling, and facility-specific approval requirements
- 130.6.1.2 Implement protocol safety validator: reject executable protocols that violate temperature, pressure, atmosphere, interlock, reagent compatibility, or personnel qualification constraints
- 130.6.1.3 Implement role- and certification-aware execution gating: only allow authorized operators to run restricted equipment or hazardous procedures
130.6.2 Review & Audit#
- 130.6.2.1 Implement preflight safety review packets: protocol summary, hazards, required controls, emergency procedures, waste plan, and approval signatures
- 130.6.2.2 Implement incident and near-miss logging tied to sample, protocol step, instrument session, and operator
- 130.6.2.3 Implement audit-ready trace export for regulated or high-consequence research programs
130.7 Closed-Loop Experiment-Compute Discovery (@kalika/closed-loop-experiments)#
130.7.1 Closed-Loop Orchestration#
- 130.7.1.1 Implement experiment-compute loop orchestrator: hypothesis -> DOE -> protocol generation -> execution -> QC -> model update -> next-experiment recommendation
- 130.7.1.2 Implement real-time model updating: assimilate new measurements into Gaussian-process, Bayesian, and neural surrogate models with calibration-aware uncertainty updates
- 130.7.1.3 Implement discrepancy modeling: learn the systematic gap between simulation and experiment as a function of process conditions, microstructure, and measurement modality
130.7.2 Domain-Specific Loops#
- 130.7.2.1 Implement battery closed-loop campaign: formulation or synthesis changes -> electrochemical test -> impedance/XRD follow-up -> updated ranking for next candidates
- 130.7.2.2 Implement catalyst closed-loop campaign: composition/support/process variation -> activity/selectivity/stability measurements -> microkinetic posterior update -> next-candidate selection
- 130.7.2.3 Implement thin-film discovery campaign: deposition parameter sweep -> optical/electrical/structural measurements -> process-window refinement -> device-oriented ranking
- 130.7.2.4 Implement structural-alloy campaign: heat-treatment or composition plan -> microscopy/mechanics measurements -> multiscale model calibration -> next process condition selection
130.8 Experimental Knowledge Capture, Validation & Integration#
130.8.1 Data Fusion#
- 130.8.1.1 Integrate all experimental outputs into the Phase 128 knowledge graph with confidence, calibration state, and full sample lineage preserved
- 130.8.1.2 Implement theory-experiment reconciliation workflows: automatically compare predicted spectra/properties/microstructures against measurements and surface disagreements that warrant new simulations or re-measurement
- 130.8.1.3 Implement publication- and patent-ready experiment packs: protocol, raw data, reduced data, QC evidence, sample lineage, and computational context bundled together
130.8.2 Validation & Benchmarks#
- 130.8.2.1 Implement DOE engine tests: verify orthogonality, balance, constrained design validity, and Bayesian acquisition behavior on synthetic objectives
- 130.8.2.2 Implement protocol execution simulations: failure injection for depleted reagents, blocked instrument access, and unsafe transition attempts
- 130.8.2.3 Implement sample lineage audit tests: no orphan samples, no untracked transformations, and no custody gaps for archived datasets
- 130.8.2.4 Implement end-to-end closed-loop benchmark: synthetic but realistic campaign where Kalika chooses experiments, updates its model, and converges to a target material/process window efficiently