Pratik Dhanave

Led the BigQuery cost optimization engagement for Tata Group — query refactoring, MERGE redesign, capacity-based slot reservation transition. Delivered ₹100 Cr+ (~$12M) in savings, 57% data warehouse cost reduction. Productized into Searce's recurring GCP managed service offering.

Kubernetes transaction platform handling 30K+ TPS for telecom/FinTech partner integrations — led 10 engineers, PCI-aligned, idempotent processing, error-code orchestration with exponential backoff and DLQ resilience.

Zero-trust voting and analytics platform for the AFL Brownlow Medal — 100K+ votes, 10K+ concurrent users on Cloud Run during live broadcasts, secured with Cloud KMS + Security Command Center.

Social platform backend serving 1M+ users — reduced API response latency 47% via protobuf contracts and service consolidation; led a remote-first international team of 4 engineers; 80%+ automated test coverage.

High-compliance Open Banking portal for UAE (ADGM/DIFC) and Saudi (SAMA) markets with consent management, data residency, and RAG-based semantic search on AlloyDB AI as a VectorDB — 37% latency reduction.

Gemini-powered BigQuery analyzer that detects SQL anti-patterns (full scans, inefficient joins, missing partition filters) and recommends optimizations — delivered 57% cost reduction (Tata engagement).

Go-based multi-agent financial assistant on Microsoft's MARA — 15 role-specialized agents (supervisor, analyzer, forecaster, anomaly_detector, recommender, llm_auditor, etc.), JWT + RBAC, AES-256-GCM envelope encryption, full OpenTelemetry tracing.

Original open-source medical AI platform on MARA — virtual physician panel inspired by Microsoft's MAI-DxO + SD-Bench. Role-specialized agents (intake, questioner, test_planner, cost_guardian, diagnostician, reasoning_verifier); FHIR R4 + HL7 v2 aware; cost-aware diagnostic budget enforcement.

Core contributor to Google's open-source Spanner migration tool — built backend APIs for the Intelligent Schema Assistant, CDC pipelines (Datastream · Pub/Sub · Dataflow), and improved post-migration query performance by 40–60% through PK design, indexing, and table interleaving.

The philosophy, architectural clarity, and vendor lock-in concerns that drove the migration. Provider abstraction as the foundation for portable agents.

How orchestration shifts from implicit callbacks (ADK) to explicit executor loops (MAF). Why this matters for observability, auditability, and control.

Porting session state from ADK's opaque state dict to MAF's explicit AgentThread. Token budgeting, long-term memory, and conversation audit trails.

Porting tools and adding governance layers (audit, approval, policy). DLP enforcement with OPA. Multi-step tool composition and error handling.

Zero-config local (Ollama), cost-optimized (OpenAI), and production (Azure Foundry) provider setup. Hybrid multi-provider orchestration.

Porting callbacks to middleware. Composable decorators for audit logging, retry logic, token enforcement, and OpenTelemetry integration.

Cloud Run deployments, agent-to-agent communication, load balancing, graceful shutdown, and production observability integration.

What worked, what was hard, and what we'd do differently. Real numbers: agent count, timeline, bugs found, provider swaps, governance policies built.

Every Professional Cloud Security Engineer exam bullet, mapped to a file path in an RBI FREE-AI aligned Go platform. Where the implementation matches, where the analog substitutes, and where the honest gaps are.

Request → N-eyes approve → window-of-time → automatic expiry, with every transition written to a hash-chained audit log. The package that closes Gap #1 from the PCSE map.

The mental model that says no two adjacent layers share a single point of failure for the same class of attack. From TLS to OTel, the eleven layers a customer request crosses before an answer comes back.

Twelve months of running multi-agent AI in a regulated context. SLIs that matter, the incident runbook, drift detection, continuous adversarial testing, secret rotation, compliance posture as code.

The long-form security narrative for a multi-agent financial assistant — authentication, authorisation, tenant isolation, dual-identity audit, envelope encryption, hash-chained logs, governance, red team, BCP.

Board policy as a YAML file the risk team owns. Annexure VI as a database query. Every governance recommendation rendered as a file path in a Go repository.

Stdlib over libraries, single binary over framework, fail-closed defaults over forgiveness. The boring-on-purpose case for choosing Go to ship a multi-agent system into a regulated environment.

Every one of the 26 RBI FREE-AI recommendations, mapped to a specific file in a working multi-agent platform. What's ✅ done, what's 🟡 partial, what's ⚪ honest gap.