Albert Dulout

Albert Dulout

Applied AI, from technical discovery to delivered systems.

Applied AI Engineer and Data Scientist based in Singapore. I work from design and implementation through evaluation, cloud delivery and handover.

Contact me Download résumé

Choose your reading depth

Recruiter view keeps project summaries concise. Technical view adds implementation details and technical context.

Selected work

Two client engagements, one internal platform and one personal project. Open a project to see the problem, my contribution and the decisions behind it.

Document intelligence Client work at PALO IT · 2025–2026 View project details

Delivered the backend, AI and cloud foundations for four short-cycle enterprise PoCs spanning regulatory comparison, loan-agreement clause matching, NDA review and batch résumé screening.

Compliance, Treasury, Legal and HR teams needed to review long PDF and Word files faster without accepting opaque model answers. The legal and comparison workflows preserved source text or page locations for human verification; the screening workflow returned structured reasons behind each candidate score.

I designed and implemented the Flask backend services and AI pipelines for ingestion, parsing, retrieval and structured comparison. Depending on the use case, the systems combined Apache Tika or Azure Document Intelligence with hierarchical chunking, BM25 and Vector Search, Semantic Reranking, schema-constrained Azure OpenAI outputs, bounded asynchronous batches, retries and PyMuPDF annotations. Colleagues built most of the React interfaces.

The four PoCs produced reviewer-ready outputs rather than generic chat responses: annotated document comparisons, covenant analysis across up to five agreements, NDA summaries and clause trackers, and ranked candidate reports with Excel exports and generated interview questions.

I also owned much of the cloud delivery foundation: multi-service Docker packaging behind Nginx, image publishing through Azure Container Registry, deployment to Azure App Service and Container Apps, Blob Storage for source and generated artifacts, Key Vault-backed configuration with Managed Identity, and GitHub Actions CI/CD.

Core stack Python · Flask · Azure OpenAI · Azure AI Search · Document Intelligence · Blob Storage · Key Vault · Docker · Nginx · App Service · Container Apps · GitHub Actions

Implementation details

Regulatory comparison used exact and fuzzy prefilters before batched LLM verification, reducing unnecessary model calls and returning red and green PDF annotations. Loan review indexed document-, section- and paragraph-level chunks in Azure AI Search, combining BM25, HNSW vector search and semantic ranking before structured GPT comparisons. NDA analysis overlapped summary, extraction, template checks and artifact uploads; résumé screening split multi-candidate PDF packets, derived evaluation criteria from job descriptions, ranked candidates and generated Excel reports.

Backend and interfaces
Python, Flask, Uvicorn and ASGI; React and TypeScript interfaces built primarily by teammates
AI and retrieval
Azure OpenAI, Azure AI Search, Embeddings, BM25, HNSW vector search, Semantic Reranking and structured outputs
Document processing and reporting
Apache Tika, Azure Document Intelligence, PyMuPDF, LibreOffice, openpyxl and hierarchical chunking
Azure and security
Blob Storage, Key Vault, DefaultAzureCredential, managed identity, Azure Container Registry, App Service and Container Apps
Delivery and operations
Docker, Nginx, GitHub Actions, health checks, bounded concurrency, retries and structured error handling
Route optimization Client work at PALO IT · 2025–2026 View project details

A route and resource-planning platform used by waste-collection operations across two countries, with configurable objectives and clear reporting when a plan is not feasible.

Planning relied on inconsistent Excel and CSV workbooks and had to account for vehicle capacities, time windows, driver shifts, repeat trips, competing objectives and incomplete road data.

I led delivery end to end: the OR-Tools solver, resilient workbook ingestion, scenario comparison, KPI and route-level reports, the React workflows and cloud deployment.

Repeat trips are modelled as shift-aware virtual vehicles. Soft penalties return unserved stops explicitly, so the planner reports when a plan is infeasible instead of failing silently. Distances can come from Haversine, OSRM or the Google route matrix, balancing road realism, cost, rate limits and offline operation.

Core stack Python · FastAPI · OR-Tools · React · OSRM and Google Routes · Azure Container Apps

Implementation details

The solver covers capacity, service time, shift, depot, unload and objective constraints, with objective modes for time, distance, fuel, carbon, cost, coverage, revenue and profit scaled to integers for OR-Tools. Around it sit tolerant workbook parsing, recursive moving-average weight forecasting, scenario comparison and operational reports.

Language and frameworks
Python, FastAPI, Uvicorn, React
Optimization and data
Google OR-Tools constraint programming, pandas, NumPy, openpyxl, pdfplumber
Geospatial
Offline Haversine baseline, cached OSRM table API, Google Routes matrix
Cloud and delivery
Docker, Azure Container Registry, Container Apps, Static Web Apps, Log Analytics
Reporting and testing
WeasyPrint PDF and XLSX reports, pytest
Resilient cloud data platform Internal platform at PALO IT · 2026 View project details

A production-deployed internal Azure data platform built around a workforce SaaS that did not provide the role-based access the business needed.

The SaaS API exposed the data, but not the access levels required by internal teams, including HR, Finance and Audit. We needed to bring that data into Azure and expose it through a controlled interface.

I designed the architecture and Azure infrastructure, then built and deployed the ingestion, Durable Functions orchestration, normalized PostgreSQL model and role-scoped API. Bicep provisions the Functions, database, Key Vault, private networking, NAT egress, monitoring and separate runtime and deployment identities; Entra ID protects the read endpoints for each audience.

I also created an MCP server so the team could access the governed platform through tool-based AI workflows.

Checkpoints and completeness checks make interrupted syncs safe to resume. They also prevent a partial import from being mistaken for deleted source data.

Core stack TypeScript · Azure Functions · Durable Functions · PostgreSQL · Drizzle · Entra ID · Bicep · MCP

Implementation details

OAuth pagination, retries, checkpoints and idempotent upserts feed a normalized PostgreSQL model across ten core sync domains. The deployment uses Functions Flex Consumption, PostgreSQL Flexible Server, private endpoints, VNet integration, NAT Gateway, managed identities, Key Vault, Application Insights, Log Analytics and GitHub Actions with OIDC. Separate identities limit privileges across infrastructure deployment, application release, schema migration and data operations.

Runtime
TypeScript, Azure Functions, Durable Functions
Data
PostgreSQL, Drizzle, role-scoped reporting views
API and access
OAuth ingestion, Entra ID, Easy Auth, role-scoped endpoints and an MCP server for team access
Azure infrastructure
Bicep, Functions Flex Consumption, PostgreSQL Flexible Server, Key Vault, managed identities, private endpoints, VNet integration and NAT Gateway
Delivery and operations
GitHub Actions OIDC, Application Insights, Log Analytics, Azure Monitor alerts, deployment smoke tests and recovery workflows
Mathilde Recipes Personal project · 2026 View project details

A family recipe platform that turns scanned documents and phone photographs into a searchable French-first collection, with English translation, a recipe-aware assistant and transparent nutrition estimates.

Visit recipes.dulalbert.com

Family recipes arrived as scanned documents and phone photographs. I wanted to turn them into something useful while learning Cloudflare and keeping a self-funded side project inexpensive.

I build and maintain all of it: the ingestion and quality pipeline, the catalog, the recipe-aware assistant, nutrition estimates, protected admin workflows, image generation jobs and the delivery automation.

A scheduled Dropbox sync picks up new source files and rebuilds the catalog. Most of the platform runs within Cloudflare's free tiers; the paid costs are the domain and on-demand image generation.

Core stack React · TypeScript · Cloudflare Pages and Functions · D1 · Vectorize · R2 · Queues

Implementation details

The pipeline scans source files, extracts text or image content, requests structured Gemini output, validates it, assigns provenance and tags, removes mirrored and fingerprint duplicates, then rebuilds a deterministic catalog index. Skip reasons, validation errors and quality fixes are recorded rather than silently published. The application adds search, cooking mode, CIQUAL-based nutrition assumptions, conditional retrieval and protected review workflows.

Frontend
React, TypeScript, Vite, i18next, custom CSS
Runtime and hosting
Cloudflare Pages, Pages Functions, Cloudflare Access on the API routes
Data and retrieval
Cloudflare D1, Vectorize, Workers AI with Qwen3 embeddings
Media and jobs
Cloudflare R2, Cloudflare Queues, sharp, heic-convert
AI
Google Gemini for extraction, classification, translation and chat; gpt-image-2 for recipe images
Ingestion
Scheduled Dropbox sync, Mammoth for DOCX, pdf-parse for PDF, CIQUAL nutrition data
Automation
GitHub Actions scheduled pipelines, Wrangler for Pages and Worker configuration

How I work

The parts of delivery I can own, from the first technical discussion to a system in use.

Understand the problem
Turn an ambiguous AI or data request into a testable technical approach, with clear constraints and tradeoffs.
Build the system
Develop the models, retrieval workflows, optimization, APIs, data pipelines and interfaces needed for a working product.
Put it into use
Deploy on Azure, automate delivery, add monitoring and design for safe recovery when something fails.
Help teams own it
Document decisions, explain the system, work with stakeholders and mentor the engineers taking it forward.
Use AI throughout delivery
Use agentic AI for technical discovery, meeting recording and synthesis, coding, testing, documentation and review across the full software development lifecycle.

Experience

Engineering school, then analytics work, then applied AI delivery.

Data Scientist, Innovation Lab

PALO IT · Singapore · March 2023–present · promoted in February 2025

I design and deliver software, data and AI systems from technical discovery through architecture, implementation, cloud deployment and handover. The work spans backend and full-stack applications, connected hardware, AI engineering, platform infrastructure and DevOps.

Across three years, I have built enterprise document-intelligence and retrieval systems, a full-stack logistics route-optimization product, an internal Azure data platform, Arduino-based hardware prototypes, and a legacy-modernization graph using Tree-sitter and Neo4j. For a public-sector exhibition, I evaluated image-generation models and inference paths against a sub-10-second response-time target, balancing latency, output quality and deployment constraints.

I also built reusable lab infrastructure for automated application delivery, including an Azure Dev Center catalog of Bicep templates for Container Apps, PostgreSQL, Static Web Apps and AI/RAG applications. Managed identities, scoped RBAC and GitHub OIDC let teams provision and deploy standardized environments without subscription-level access.

For two .NET modernization workstreams, I designed and delivered Azure infrastructure and deployment foundations rather than application features. For RewardsManagement, the target architecture ran Dockerized .NET 8 services on Container Apps behind Application Gateway and WAF, with private VNets and endpoints, Azure SQL, Azure Container Registry, Key Vault and managed identities. Across the workstreams, I also supported Azure DevOps CI/CD, health checks, database initialization and migrations, repeatable migration data, private connectivity and release coordination.

In presales, I translate client pain points into technical approaches and proposals, help size delivery teams and infrastructure, and present solution options, constraints and trade-offs to prospective clients.

Build
Python, TypeScript, Node.js, Flask, FastAPI, React, Arduino
AI and data
Azure OpenAI, Azure AI Search, embeddings, RAG and semantic reranking, Replicate, LoRA fine-tuning, OR-Tools optimization, Computer Vision, PostgreSQL, pandas
Cloud and delivery
Azure Functions, Azure Dev Center, Container Apps, App Service, Bicep, Docker, Key Vault, managed identity, GitHub Actions, Azure DevOps
Working with teams
Technical discovery, solution architecture, presales proposals, team and infrastructure sizing, client presentations, documentation and mentoring

Analytics Engineer Intern

Papernest · Barcelona, Spain · February–August 2022

I worked with the Southern Europe growth team on marketing analytics and built campaign-improvement tooling on the Google Ads and Sheets APIs. I increased Google Ads conversion imports by 6% by tracing and restoring records dropped in the SQL and ETL flow. The more complete conversion signal improved campaign relevance and helped lower cost per acquisition.

I also automated the team's Slack alerting so anomalies in campaign performance surfaced without anyone running a query, and I maintained the reporting the team used for weekly spend decisions.

Tools
SQL, BigQuery, Python, pandas, Google Ads API, Google Sheets API, Slack API, JavaScript, Data Studio

Master Data Analyst Intern

Nestlé Europe · Paris, France · June–December 2021

I built a pricing simulator in Excel and VBA on top of data extracted from SAP and SQL, so the commercial team could test price scenarios without going back to the data team each time.

I automated several recurring data-management and analysis tasks, and evaluated the Alation data-governance platform with counterparts at Purina in the United States.

Tools
SAP, SQL, Excel and VBA, Alation

Education and credentials

Engineering Degree / MSc, Data Science and Optimization

IMT Atlantique · Brest and Nantes, France · September 2019–April 2023

Coursework in machine learning, Computer Vision, natural language processing, data mining, time-series forecasting, optimization algorithms, big-data architecture and data visualization, plus a blockchain application built on Avalanche.

Preparatory Class for the Grandes Écoles

Lycée Montaigne · Bordeaux, France · September 2016–June 2019

Two years of intensive sciences with mathematics as the major and physics and chemistry as minors, the competitive entry route into the French engineering schools.

  • Microsoft Certified: Azure AI Engineer Associate
  • Google Data Analytics Professional Certificate
  • IELTS Academic, Band 8.0 / C2
  • French (native), English (professional), Spanish (conversational), Russian (intermediate)

Beyond the work

I am French and based in Singapore. Outside work, I am usually cooking, training or learning something new.

Cook
Family recipes, often with Mathilde Recipes open beside the stove.
Train
Weightlifting, long-distance running and rugby.
Compete
Two-time Singapore Development League rugby champion.
Learn
Learning another language.