BoardDeck Sync: Multi Source KPI to Slide Auditor
by Petrica Radan
Reconcile CRM, Stripe and ops exports first—blocking executive Google Slides updates automatically whenever cross-system KPIs disagree.
Data Science & Analytics leader delivering production machine learning and Generative AI systems in large-scale retail. Currently lead a team of data scientists across demand forecasting, promotional analytics, and agentic AI. Deep, hands- on command of the Google Cloud stack (Vertex AI, BigQuery/BQML, Cloud Run, Dataform) paired with rigorous statistical and causal-inference methods. Known for converting ambiguous business problems into deployed, production-grade systems — strengthening inventory and replenishment planning, sharpening commercial and promotional decisions, and measuring the true causal impact of campaigns — rather than one-off prototypes.
by Petrica Radan
Reconcile CRM, Stripe and ops exports first—blocking executive Google Slides updates automatically whenever cross-system KPIs disagree.
by Petrica Radan
Catch competitor stealth seat-minimum hikes, paywalled features and Shopify SKU review defect spikes—with capped Apify scrapes and zero VPS upkeep.
by Petrica Radan
Bridge HubSpot or Zendesk tickets to Linear, GitHub and Slack with timezone business-hour SLA clocks, bug deduplication and ARR blast-radius scoring.
by Petrica Radan
Verify every sales commission cent before payroll: marginal accelerators, split credits, pro-rated quotas and 90-day clawbacks traced deal by deal.
by Petrica Radan
Cut BigQuery scan and storage waste using read-only INFORMATION_SCHEMA metadata—with ranked dollar attribution and ready-to-review partitioning SQL.
by Petrica Radan
Track how ChatGPT, Perplexity and Google AI Overviews cite your brand vs competitors over time—with 0% URL hallucination and no $800/mo CRM lock-in.
by Petrica Radan
Prevent 5x surprise Apify, proxy and LLM bills by auditing true cost-per-valid-result and tripping circuit-breakers on >20% bot-block loops.
by Petrica Radan
Block silent prompt, model and tool-schema regressions in CI with deterministic Google ADK trajectory checks, injection probes and SHA-256 provenance.
by Petrica Radan
Catch duplicate invoices, PO price creep, short deliveries, BEC bank swaps and VAT math errors before your ACH batch goes out.
by Petrica Radan
End month-end spreadsheet archaeology: deterministically reconcile bank statements, Stripe payouts and ERP ledgers locally in Python Decimal math.
by Petrica Radan
Convert your official syllabus, slides and notes into a citation-locked quiz and SM-2 spaced-repetition coach that never wanders off-curriculum.
by Petrica Radan
Turn a year of messy receipts, platform payout CSVs, home-improvement invoices and appliance warranties into deduplicated ledgers and archive names.
by Petrica Radan
Inspect suspicious texts, spoofed invoices and breach notices with zero-password forensics, then lock down your critical accounts in 30 minutes.
by Petrica Radan
Turn exported bank CSVs into a multi-tab cashflow, BNPL, sinking-fund and Snowball-vs-Avalanche debt workbook you permanently own.
by Petrica Radan
Cut through sponsored rankings and expose hidden refill, energy and warranty traps with an exact 1-year and 3-year Total Cost of Ownership shortlist.
by Petrica Radan
Tailor ATS-clean resumes, cover letters and LinkedIn copy for non-coding roles — backed by a deterministic audit proving zero invented credentials or metrics.
by Petrica Radan
Run a 40-point risk screen on leases, renovation bids, auto sheets, gym contracts and move-out deductions — with verbatim page citations and polite counter-language.
by Petrica Radan
Turn cancelled flights, missed connections, delayed bags and multi-currency hotel or meal receipts into a timezone-verified reimbursement packet.
by Petrica Radan
Find forgotten subscriptions, stealth price hikes and scattered Buy Now, Pay Later installments from exported CSVs — with zero bank logins and zero success fees.
by Petrica Radan
Reconcile hospital, ER and outpatient bills line by line against your insurer EOB before you pay a cent you don't owe.
by Petrica Radan
Design an autonomous SRE agent that diagnoses failing jobs, writes a minimal fix in a throwaway sandbox, proves it through quality, security and risk gates, and opens a pull request, never writing to production directly.
by Petrica Radan
Run an evidence-based go / no-go launch review of an LLM agent: a weighted 32-check rubric across design, tools, security, evaluation, reliability, cost and operations, with a scored report, blockers, conditions and owners.
by Petrica Radan
Instrument LLM agents so every run can be traced, costed and debugged: OpenTelemetry GenAI spans, a structured event log, the metrics and alerts that matter (cost per successful run, cache hit ratio, tool errors), dashboards and BigQuery queries.
by Petrica Radan
Make your agent's run loop robust: every stop reason handled, hard budgets on steps, tokens and time, classified retries, per-tool timeouts, idempotent writes, checkpoint-and-resume and fallbacks, with a tested reference loop.
by Petrica Radan
Engineer what goes into your agent's context window: system prompt structure, cache-friendly ordering, retrieval, memory and compaction, token budgets, and the cost levers in the right order, plus a cache-buster scanner.
by Petrica Radan
Design the tools your LLM agent calls so it picks the right one with valid arguments first time: granularity, names, descriptions, schemas, results, errors and side-effect classes, plus a linter for tool definitions.
by Petrica Radan
Produce vertical Shorts with AI video models the way commercials are made: a two-shot board with one action per shot, critic review rounds, first frames, model-watched dailies with retakes, typeset captions, music, 9:16 delivery and YouTube upload.
by Petrica Radan
Turn a video's comments into decisions: unanswered viewer questions, content gaps, sentiment, feedback themes, loyal commenters, evidence-backed video ideas and reply drafts for you to approve.
by Petrica Radan
Produce long-form articles that are researched, cited and on-topic: staged pipeline, source credibility scoring, evidence-pack drafting, automatic checks for unsourced numbers and topic drift, and WordPress publishing.
by Petrica Radan
Run a B2B LinkedIn company page with an AI agent on the official API: grounded posts, pillar rotation and funnel stages in code, safe replies, hard limits, dry-run mode, post statistics and data-driven boost picks.
by Petrica Radan
Ship Google ADK agents and multi-agent systems to Cloud Run without the classic failures: missing root_agent, 403s to private MCP servers, lost sessions, CORS *, and GPUs you never needed.
by Petrica Radan
Build voice-first apps on the Gemini Live API where speech drives small, validated edits instead of full redraws, with the API key kept out of the browser and sessions that survive disconnects.
by Petrica Radan
Document, profile and govern BigQuery so people and AI agents can find and trust data: sync that never overwrites human docs, schema-change alerts, safe profiling, PII detection, usage, lineage and AI-drafted descriptions.
by Petrica Radan
Turn "executive sales dashboard with revenue trend and store ranking" into a real Looker Studio report on BigQuery, without the LLM ever inventing a field.
by Petrica Radan
Build an agent in Google Chat that reads mail, schedules meetings and searches Drive as the person asking: request verification, async replies, per-user OAuth, Workspace MCP and safe deployment on Cloud Run.
by Petrica Radan
Know whether an agent or prompt change is actually better: golden datasets, calibrated LLM judges, confidence intervals, paired comparisons, power analysis and CI regression gates.
by Petrica Radan
Build, test, secure (OAuth 2.1, multi-tenant) and deploy MCP servers that Claude, ChatGPT and IDEs can depend on, with a server template and a protocol smoke test.
by Petrica Radan
Build a "chat with your data" agent that never runs dangerous queries, never surprises you with a warehouse bill, and never states a wrong number with confidence.