Software Engineer · Generative AI · Available for projects
I'm Leon Achata, a software engineer specializing in generative AI. I design and build enterprise GenAI systems for Apple through TCS. For companies in the US, I deliver agents, RAG assistants and document pipelines end to end — architecture, backend, interface and deployment on your cloud — and hand over clean, documented code.
Based in Montevideo (UTC−3) — my day overlaps the full US workday.
Every step traced, validated and priced — that's the standard I build to.
Where I've built AI
Apple
via TCS
JLR Analytics
AI consulting
UPCH
Research university
Cardiomed
Healthcare
Services
Four kinds of projects where language models pay for themselves quickly. Each one starts with a free discovery call and a fixed-scope proof of concept on your real data.
Invoices, IDs, contracts and forms turned into validated, structured data — straight into Excel, your database or your ERP.
What you get
A chat assistant that answers from your own documents and shows where each answer came from. Private, searchable and measurable.
What you get
Multi-step agents that connect your tools — email, Slack, ticketing, internal APIs — and take repetitive work off your team's plate.
What you get
Let non-technical people query your databases in plain English, with guardrails so the model can never run unsafe SQL.
What you get
Selected work
Open-source builds, each one a working version of a problem companies pay to solve. Every project links to its code.
Invoice PDFs in, validated structured data out — ready for Excel, PostgreSQL or an ERP.
Agents that share one gateway to Bedrock, OpenAI and Gemini — with caching, cost tracking and MCP tools.
Describe your stops in plain language and get an optimized route with turn-by-turn directions.
A conversational real-estate assistant that turns chat into safe, parameterized SQL.
Chat → Extract filters → Validate → Build SQL → Top 5 matches
A document Q&A system on AWS, deployed end to end as infrastructure as code.
S3 upload → Ingestion Lambda → Titan embeddings → pgvector → Claude on Bedrock
IDs, driver's licenses and passports read and structured automatically to pre-fill forms.
Image → Validation → OCR → LLM field mapping → Form data
An MCP server that gives any agent access to 15+ Amazon Bedrock models through one interface.
Agent → MCP (stdio / SSE) → Router + cache → Bedrock models
A portable ECG Holter on an ESP32 that uploads its recordings to AWS on its own.
ECG + IMU capture → SD storage → MQTT request → Presigned URL → S3 + Lambda
Process
30 minutes. You walk me through the problem; I ask about volume, systems and constraints. You leave with an honest answer on whether AI is the right tool.
One to two weeks on your real data. A working version you can click, plus the numbers that matter: accuracy, cost per run, latency.
Hardening, integrations, evaluation and deployment on your cloud. Weekly demos, no black boxes.
Documentation, code in your repository and a support window after launch. You own everything.
Stack
GenAI & agents
Backend & data
Cloud & ops
Frontend & product
Engineering practice
Experience
Mar 2026 — Present
Design and build enterprise GenAI automation for internal workflows: LLM pipelines, RAG and agent orchestration integrated with collaboration tools, taken from proof of concept to production. Daily work with engineering teams across the US and APAC.
Sep 2025 — Feb 2026
Built multi-agent systems with LangChain, LangGraph and MCP, deployed on AWS (EKS, ECR, Bedrock, S3, DynamoDB, CloudWatch). RAG with pgvector and local models, document chatbots with OCR and speech-to-text, and agent orchestration for an AI-driven robotic arm for POS systems.
Apr 2025 — Sep 2025
Agents and conversational workflows for research teams; RAG systems with FAISS and ChromaDB; reproducible pipelines for text, image and structured data.
Oct 2023 — Mar 2025
Transformer models for medical text classification and information extraction, LangChain assistants for internal medical queries and automated technical reports, and Python workflows for data analysis.
Jan 2020 — Oct 2023
Where it started: data analysis, reporting and the first machine learning models.
2026 — 2029
B.Sc. Software Engineering
Universidad Católica del Uruguay
2021 — 2025
B.Sc. Biomedical Engineering
Pontificia Universidad Católica del Perú
About
I'm a software engineer specializing in generative AI, with six years building data and AI systems — from analytics pipelines to multi-agent platforms running on AWS. The model is rarely the hard part. The engineering around it is: APIs, data flow, infrastructure, evaluation and observability.
Today I design and build enterprise GenAI systems for Apple through TCS, taking them from proof of concept to production alongside engineering teams in the US and APAC, entirely in English.
With clients I work end to end — architecture, backend, the interface people actually use, and deployment. One senior engineer accountable for the whole system, with clear scope and a clean handover.
LA
Software Engineer · GenAI
FAQ
After the discovery call I send a fixed-scope proposal for the proof of concept, so you know the cost before any work starts. Production work can be fixed-scope or a monthly retainer — whatever fits the project.
I'm used to enterprise security requirements. I can sign an NDA, work inside your own cloud account, and use models that don't train on your data (AWS Bedrock, Azure OpenAI or self-hosted open-source models).
Whichever fits your cost, privacy and quality needs: OpenAI, Anthropic Claude, models on AWS Bedrock, or open-source models. During the proof of concept I show you the trade-offs measured on your own data.
You do. The code lives in your repository, with documentation and a handover session so your team can run and extend it.
I take on a small number of freelance projects at a time so each one gets focused attention. Tell me your timeline on the call and I'll be straightforward about whether I can meet it.
Contact
Tell me about it. The first call is free, and you'll come away with a clear idea of what's possible and what it would take.
Book a 30-min callOr write to
leonyemin@gmail.com