Skip to main content
Open to senior engineering and tech lead roles

Marcelino Sandroni DiasSoftware engineering that moves revenue and scale.

Senior Software Engineer & Tech Lead

  • R$ 24M/yrRevenue protectedannual contract value rescued from cancellation
  • 100M msgs/dayScale handledtelemetry events ingested across the pipeline
  • 15minDeploy, from 4hdeploy duration after the delivery pipeline

Six years writing software (2021–2026), on top of fifteen years in corporate financial governance (2005–2020). I work end to end across architecture, backend, frontend and DevOps — Go, distributed systems, ClickHouse, Kafka and Kubernetes at fintech scale — and I lead the teams that ship it. The accounting background is not a detour: it is why I can price compute cost and operational risk as a balance-sheet liability instead of guessing.

// ID: MSD-01
Portrait of Marcelino Sandroni Dias, Tech Lead and Software Architect
100% ACCURACY

// TECH LEAD & SOFTWARE ARCHITECT

Go · .NET · Java · Kafka

[NOMINAL]
// STACK: FULL
LATENCY < 0.1ms
THROUGHPUT: 100M MSGS/DAY
100% SLA COMPLIANT
// IMPACT

What the engineering was worth

Capital protected, scale absorbed, and the operational cost of shipping removed. Every figure below is on the resume with the detail behind it.

COMBINED EXPERIENCE

21 years

15 years of corporate financial governance (2005–2020) plus 6 years of software engineering (2021–2026). The accounting background is what lets me price compute cost and operational risk as balance-sheet liability.

MONTH-END30→5 DAYS (-83%)
BACKEND CORE

100M+

Telemetry messages processed per day across distributed pipelines, with Apache Kafka partitioning, ClickHouse columnar storage and p99 edge latency under 10ms.

PROTECTED REVENUER$ 24M/YEAR
DEVOPS & CLOUD

-95%

Deploy time cut from 4 hours to 15 minutes, with automatic rollback in under 2 minutes, 80% fewer production incidents and infrastructure as code across AWS, Azure and GCP.

DEPLOY4H → 15MIN
ARTIFICIAL INTELLIGENCE

100%

100% accuracy on license-plate recognition with computer vision, alongside enterprise RAG, pgvector similarity search and TensorRT-optimized real-time video inference.

RECALL100% PROD
// TECH STACK

Four disciplines, one decision per problem

Every technology listed here appears on the resume, in the category it appears in here.

Frontend & UI

Enterprise microfrontends, accessible design systems and high-frequency rendering with no performance regression.

ReactNext.js (RSC, SSR)AngularFlutterReact NativeTailwind CSSDesign SystemStorybook

Backend Core

Hexagonal architecture, domain-driven design, high-throughput messaging and both relational and columnar data modeling.

GoJava (Spring Boot)C# (.NET Core)Node.js (NestJS)Apache KafkaPostgreSQLClickHouseRedis

DevOps & Cloud

Infrastructure as code, Kubernetes orchestration, distributed observability and continuous delivery with quality gates.

DockerK8s (Helm, HPA)IaC (Terraform)AWSAzureGCPOpenTelemetryCI/CD

Artificial Intelligence

GenAI applied to the development cycle, context-augmented retrieval and real-time edge inference.

LLM Integration (OpenAI, Claude, Gemini)RAGpgvectorQdrantTensorRTLangChainAutonomous Agents
// TRACK RECORD

Track Record: Production Systems Experience

Every role, with what it changed and who it was for.

Jan/2026 – Present // Remote

DGT Tecnologia

Senior Software Engineer & Tech Lead

IMPACT: R$ 24M/YEAR SAVED
  • Rescued a R$ 24M/year (R$ 2M/month) contract in Florianópolis, lifting license-plate recognition accuracy from <60% to 100% and removing the cancellation risk.

  • Multiplied throughput 10x on 100M messages/day by migrating MySQL to ClickHouse with Databricks ETL and Go and .NET microservices.

  • Led 10 people (Dev, QA, BA, PO, PM) and cut development time 40% through a design system and shared libraries.

Dec/2022 – Dec/2025 // Remote

Antlia

Tech Lead & Strategic Consultant

IMPACT: +45% ASSETS UNDER MGMT
  • Grew assets under management 45% in 1 year, with settlement moving from D+1 batch to real time and uptime from 95% to 100%.

  • Created opportunities on the order of R$ 50 million and automated manual processes, saving R$ 500 thousand per year.

  • Led 8 developers (3 principal) modernizing COBOL, VB and C# onto Java Spring and Angular, with hexagonal architecture, Kafka and 95%+ test coverage.

Mar/2022 – Nov/2022 // Hybrid

Banco Itaú

Full Software Engineer & Tech Lead

SCALE: +R$ 100B UNDER CUSTODY
  • Designed the asset-management platform operating +R$ 100 billion under custody, with B3 integration and a unified real-time view.

  • Rescued a critical launch hit by infrastructure failure, fixing the messaging microservices in under 24h.

  • Implemented automated testing across 3 layers (.NET, Angular, Flutter), raising code quality 8x with 80%+ coverage on a 9-person team.

Jun/2021 – Jan/2022 // Remote

Pollux Technologies

Full Software Engineer (Promoted in 3 months)

CLOUD SAVINGS: -50% INFRA COSTS
  • Led an AWS public-cloud migration delivered in 3 months against a 6-month plan, with 2x performance and 30% cost reduction.

  • Cut a client's monthly AWS spend 50% through invoice auditing, autoscaling and moving workloads to spot instances.

  • Fixed messaging failures and data-integrity defects in healthcare, taking record accuracy to 100% and appointment predictability from 0% to 40%.

Jan/2005 – Dec/2020 // São Paulo

Accounting Consulting Firms

Accounting Analyst & Manager

MONTH-END: 30 DAYS → 5 DAYS (-83%)
  • Cut the accounting close from 30 days to 5 days (−83%), freeing 20+ people for higher-value work.

  • Saved up to R$ 1 million per year in taxes through advanced tax planning and strategic accounting reclassification.

  • Allocated R$ 2 million into high-yield investments (12–15% p.a.) and eliminated 95% of manual reconciliation errors.

// HOW I LEAD

The part that is not a project

A tech lead is judged on decisions and people, not on commits. These are the decisions and the people, with what came out of them.

Crises, first

A production problem is a prioritization problem before it is a technical one. I map the defects, rank them by business impact, and put the team on the critical path within hours.

Took over a public-sector contract that was weeks from cancellation, mapped 47 critical defects in 48 hours, and shipped daily hotfixes to production until the accuracy gap closed.

Decisions in the open

Architecture goes through pull requests and written rationale, not hallway agreement. When I overrule the room I write down why, so the next person can disagree with the reason instead of the outcome.

Moved a monolith onto 12 Go and .NET services with a columnar store behind it, and the split became the reference pattern other teams adopted.

Leave the team faster than you found it

Anything solved twice gets written down once. Shared libraries, a design system and real onboarding docs are how a team stops being one person's memory.

Onboarding went from 3 months to 3 weeks, new features 40% faster, and 50+ shared components across 15+ projects.

Translate for the room that pays

I can defend a cost decision in the language of the balance sheet, not only in the language of the codebase. Fifteen years of financial governance is what makes that a habit rather than a translation.

A MySQL columnar migration turned a 30s dashboard query into 200ms and cut infrastructure load 70% — reported to leadership as time and storage, not as a benchmark.

// TECHNICAL BLOG

Whitepapers and field notes

Read all articles
// CONTACT

Let's talk about your hardest problem.

Describe the problem on WhatsApp and we will work it from there. If you would rather write, the brief below is already structured for you.

Within 1 business day
// DIRECT CONTACT CHANNEL
Message on WhatsAppOpen the email brief

WhatsApp opens the conversation with the context already suggested; email opens a filled-in draft. Nothing is sent from this page and there is no form to submit.

WHATSAPP OR EMAIL // RESPONSE WITHIN 1 BUSINESS DAY