OPERATING
MODES
Five purpose-built intelligence modes, each designed for a distinct phase of the forensic workflow. Together they form the complete Forensic Intelligence Engine.
Forensic Auditor
Deterministic Compliance Audit
The Auditor. Uncompromising deterministic validation that does not care about your feelings — it cares about the math. Built for the EVMS validation team preparing an IPMDAR submission, the DCMA surveillance reviewer needing reproducible findings, and the program office that has been burned by AI-inferred verdicts on data that wasn't structurally sound to begin with. Forensic Auditor runs every DCMA 14-Point metric and EIA-748 compliance check in pure deterministic code against your uploaded schedule and cost data. No probabilistic scoring. No AI verdicts. Every metric calculated mathematically — pass or fail — with the specific task ID, activity, or cost account surfaced alongside every finding.
The forensic workflow executes in three discrete stages. Scroll to explore each capability.
Pure Math, No Inference
Every check is calculated in code. The same input always produces the same output. There is no model temperature, no probabilistic scoring, no language model deciding what "looks right." Forensic Auditor does not issue an AI verdict on data that has not passed structural validation. If a schedule violates DCMA logic checks, the engine refuses to run downstream AI analysis until the structural foundation is sound.
- ✓Binary pass/fail findings — no ambiguity, no AI guesswork
- ✓Task-level and cost-account-level traceability on every finding
- ✓Same input, same output, every time — fully reproducible
- ✓No AI verdict issued on data that has not passed structural validation
Hunting for Artificial Optimization
The engine evaluates the schedule for true physical realism — not just basic errors. It hunts for forced constraints masking schedule pressure, artificially flattened resource curves indicating machine-generated planning, broken network logic dressed up to pass surface checks, and the structural footprints of contractor manipulation that survive a casual review. Grounded in DCMA 14-Point, DOE PM-30 DIQs, GAO-20-195G Cost Estimating Guide, and NDIA PASEG schedule excellence standards.
- ✓Forced constraints masking schedule pressure detected
- ✓Artificially flattened resource curves flagged
- ✓Broken network logic dressed up to pass surface checks identified
- ✓Grounded in DCMA 14-Point, DOE PM-30, GAO-20-195G, EIA-748, and NDIA PASEG
IPMDAR-Ready Diagnostic Reports
Output is a structured, hard-hitting diagnostic report that identifies critical failures, names exactly which federal guideline was violated, and provides click-by-click Primavera P6 instructions to fix the issue at the source. Every finding is evidence-backed, traceable, and exportable as a legally defensible audit document. Supports Oracle Primavera P6 (XER), Microsoft Project (MPP, XML), Deltek Cobra (ZIP export), Asta Powerproject, SAP PS, PDF contracts, and CSV cost data.
- ✓P6 (XER), MS Project (MPP/XML), Deltek Cobra (ZIP), Asta Powerproject, SAP PS
- ✓PDF contracts and specifications, CSV cost data — no format conversion required
- ✓IPMDAR-ready diagnostic reports — legally defensible and exportable
- ✓Every finding names the exact federal guideline violated with source traceability
Guided Remediation
Turn-by-Turn Repair Sequences
The Senior Mentor. Conversational coaching, best-practice education, and click-path navigation that turns junior analysts into capable practitioners — one finding at a time. Built for the junior analyst staring at a flagged anomaly with no idea where to start, the new CAM trying to understand why their EVMS submission was rejected, or the project controls specialist who needs to fix the issue and learn enough to prevent it next time. Guided Remediation acts as the ultimate Senior Project Controls Manager — patient, supportive, technically rigorous, and grounded in actual industry-standard software manuals rather than generic AI guesswork.
The forensic workflow executes in three discrete stages. Scroll to explore each capability.
The "Why" Explained in Plain English
Every remediation session begins with a relatable analogy that explains why the metric matters before showing how to fix the underlying issue. A negative float warning isn't a number — it's a promise the project has already broken to its end date. An LOE task with discrete successors isn't a configuration error — it's a critical path that's mathematically meaningless. The analyst learns the concept, not just the keystrokes.
- ✓Conversational coaching grounded in real industry-standard software manuals
- ✓Plain-English "Why" before the technical "How"
- ✓Relatable analogies for every metric and anomaly type
- ✓Designed to leave the analyst smarter than they were five minutes ago
Click-Path Solutions in the Tool You're Already Using
Specific, bolded, step-by-step keystrokes to execute the fix in the source platform. Oracle Primavera P6 menu navigation. Microsoft Project ribbon commands. Deltek Cobra screen-by-screen guidance. The instruction set is tailored to the specific UI of the platform in use — no translation layer, no generic scheduling abstraction.
- ✓Oracle Primavera P6 — menu navigation and activity-level repair sequences
- ✓Microsoft Project — ribbon commands and schedule option configuration
- ✓Deltek Cobra — screen-by-screen cost account guidance
- ✓Asta Powerproject and SAP PS — native platform instruction sets
Anti-Hallucination Guardrails
Guided Remediation is grounded exclusively in industry-standard software documentation, EIA-748 guidelines, GAO best practices, and PASEG/PMG schedule and cost methodologies. The engine refuses to teach bad habits. If a repair path would violate a governing standard, the system flags it rather than executing it. Trust is the foundation of mentorship — and the foundation of Peveka.
- ✓Grounded in EIA-748 guidelines, GAO best practices, and PASEG/PMG methodologies
- ✓Platform software documentation used as source — not general AI training data
- ✓Repair paths that violate governing standards are flagged, not executed
- ✓Anti-hallucination guardrails enforce standards on every recommendation
Tactical Execution
Automated Project Correspondence
The Field Medic. On-demand troubleshooting and regulatory ammunition for the field. Built for the executive in the field — when a contractor pushes back on out-of-sequence steel erection, when a CAM needs to reject a bad time extension request before close of business, when a superintendent needs to decide between Retained Logic and Progress Override and doesn't have time to look it up. Tactical Execution turns dense regulatory language into plain-English, field-ready reality and delivers it in the moment of decision.
The forensic workflow executes in three discrete stages. Scroll to explore each capability.
Targeted Interrogation
Upload a specific fragnet or ask a direct question — Tactical Execution retrieves exactly the governing rule needed and nothing else. Targeted RAG pulls the specific EIA-748 guideline, GAO chapter, PASEG section, or contract clause relevant to the situation and delivers it in plain English, in the moment of decision. No 40-page manuals. No document dump. Just the citation and what it means.
- ✓Targeted RAG retrieval — only the specific rule needed, not a document dump
- ✓Plain-English translation of EIA-748, GAO-20-195G, PASEG, and applicable contract clauses
- ✓Fragnet upload or direct question — the engine pulls the governing citation
- ✓Answer delivered in the moment of decision — no manual lookup required
Software-Specific Triage
Once the rule is identified, Tactical Execution generates step-by-step repair instructions in the platform currently in use. A P6 question gets P6 menu navigation. A Cobra question gets Cobra-specific guidance. Instructions reference the exact UI, ribbon commands, and dialog boxes of the tool in the field team's hands — not a general scheduling abstraction that requires interpretation.
- ✓Oracle Primavera P6 — Retained Logic vs. Progress Override, calendar exceptions, resource leveling
- ✓Microsoft Project — Schedule Options, constraint types, baseline comparisons
- ✓Deltek Cobra — cost account setup, EAC adjustments, calendar exceptions
- ✓Asta Powerproject and SAP PS — native platform guidance
Ammunition Generation
Drafts hostile or defensive communications — RFI responses, time extension rejections, variance reports, corrective action notices, cure letters, differing site condition notices — with the governing DCMA, GAO, or contract clause citation explicitly attached. Every output is legally defensible the moment it leaves the field team's hands. Peveka isn't writing letters — it's arming a PM to win a contractual fight.
- ✓RFI responses and time extension rejections with governing citation attached
- ✓Variance reports, Corrective Action Notices (CANs), and cure letters
- ✓Differing Site Condition notices with applicable clause mapped
- ✓Every output legally defensible — designed for the field, not the boardroom
Master Strategist
Program-Level Forensic Synthesis
The Program Synthesizer. Zero-prompt program-level audits that surface material weaknesses, billing fraud, and schedule manipulation. Built for executives, owners' representatives, and program directors who need the truth about a contractor's monthly deliverables before they sign off on a progress payment. Master Strategist runs the full Tri-Node Protocol — Regulatory, Technical, and Financial — across the entire submission package simultaneously, holding EIA-748 guidelines, GAO best practices, PASEG schedule quality standards, and the contractor's raw data in active memory at the same time. It looks for the Time Machine Paradox — the moment when the contractor's narrative doesn't match the math.
The forensic workflow executes in three discrete stages. Scroll to explore each capability.
Drag-and-Drop Ingestion
Drop an entire month's update into the interface — Primavera P6 .xer, Cobra .csv, IPMR or CPR .pdf, baseline change documents, narrative reports — all at once. Master Strategist parses every file in parallel and builds a unified program model in seconds. No template to fill in, no upload wizard to navigate, no data mapping required.
- ✓P6 .xer, Cobra .csv, IPMR/CPR .pdf, baseline change documents, narratives
- ✓All files parsed simultaneously — unified program model built in seconds
- ✓No data mapping, no template, no upload wizard required
- ✓Multi-modal: structured data, PDFs, and narrative documents ingested together
Autonomous Triangulation
Without a single typed prompt, the engine cross-references every file against every other: does the financial progress claimed in the CPR mathematically exist in the schedule telemetry? Does the CAM's stated EAC fall within the IEAC confidence band the data actually supports? The Tri-Node Protocol — Regulatory, Technical, and Financial — runs simultaneously. The engine detects the Time Machine Paradox: the moment the contractor's narrative stops matching the math.
- ✓Tri-Node Protocol — Regulatory, Technical, and Financial cross-validation simultaneously
- ✓Time Machine Paradox detection — narrative vs. mathematical reality reconciliation
- ✓Zero typed prompts required — the audit runs autonomously on upload
- ✓CPR-to-schedule telemetry reconciliation and IEAC confidence band analysis
Forensic Verdict
Master Strategist produces a single executive output: a high-level forensic finding that identifies critical disconnects, names the affected control accounts and schedule activities, and automatically drafts the recommended regulatory action — Level III Corrective Action Requests, recommendations to halt progress payments, baseline change rejections, or formal cure notices. Every finding is attached to its governing citation: EIA-748 guideline, GAO Cost Estimating chapter, PASEG section, or contract clause.
- ✓SPI, CPI, TCPI, VAC, EAC, and four-formula IEAC confidence band across the full program record
- ✓Level III Corrective Action Request drafting with governing citation attached
- ✓Progress payment halt recommendations and baseline change rejections generated automatically
- ✓Designed for owners, program directors, and executives — not analysts
PARS DIQ
Mathematical Integrity Gate
The Lie-Detector. The uncompromising mathematical foundation of the engine — structural and mathematical integrity validated before any AI interpretation runs. Built as the deterministic gate every other mode depends on. PARS DIQ — the Project Assessment and Reporting System Data Integrity Qualifier — is a hard-coded implementation of the exact DIQ tests utilized by the Department of Energy and DCMA EVMS compliance teams. This is not a heuristic, not a probabilistic check, and not an AI inference layer. It is the federal compliance team's own ruleset, implemented in code, run against your raw schedule and cost data the moment a file is uploaded.
The forensic workflow executes in three discrete stages. Scroll to explore each capability.
Raw Telemetry Extraction
The moment a .xer, .mpp, .xml, or .csv file is uploaded, PARS DIQ scans every row, every task, every logic relationship, every resource assignment. No sampling, no statistical extrapolation — every record is evaluated against every applicable rule. The ruleset is the exact DIQ implementation utilized by DOE and DCMA EVMS compliance teams, not a proprietary approximation.
- ✓Implements the exact DIQ tests utilized by DOE and DCMA EVMS compliance teams
- ✓No sampling — every record evaluated against every applicable rule
- ✓Supports .xer, .mpp, .xml, and .csv formats — no format conversion required
- ✓Binary pass/fail per check — no probabilistic scoring, no AI verdict
Rigid Threshold Testing
The engine runs raw data through strict compliance filters and mathematically hunts for showstopper errors that defeat AI inference if left undetected — Missing Predecessors (DS05_001), Ghost Work (tasks with duration greater than 44 days), Hard Constraint violations (CS_MSO, CS_MFO codes), Negative Lag relationships, and WBS Traps (costs loaded on summary levels rather than discrete control accounts). Each check produces a binary pass or fail with the exact task ID, activity ID, or constraint that triggered the failure.
- ✓Missing Predecessors (DS05_001) — open-ended logic detected and flagged
- ✓Ghost Work — tasks with duration greater than 44 days identified
- ✓Hard Constraint violations — CS_MSO and CS_MFO codes detected
- ✓Negative Lag relationships and WBS Traps surfaced with task-level evidence
The Hard Gate
PARS DIQ is not a warning system. It is a structural firewall. If a schedule fails integrity validation, no AI mode — Tactical Execution, Master Strategist, Guided Remediation — processes that schedule until structural failures are resolved. The gate is not bypassable. AI inference on corrupt data produces findings that look authoritative but are mathematically unsound. PARS DIQ refuses to let that happen.
- ✓Hard gate to all downstream AI modes — not bypassable at any permission level
- ✓Tactical Execution, Master Strategist, and Guided Remediation blocked on failure
- ✓AI inference on corrupt data prevented before it can produce bad findings
- ✓Output: clean, undeniable math handed to the AI modes to build arguments on