# Luis Lacerda

> AI Systems Builder · AI Training Data and LLM Evaluation · Full-Stack Developer (TypeScript, Python), based in Florida, United States. I build software systems with AI coding agents as the workforce and myself as architect, specifier and final reviewer. Every change I ship has been written to a specification I drafted, tested against it, and approved by me through a pull-request workflow.

## Experience

### Founder, MTOS / ABBYCRM and TFT Legal Services — 2026 – Present
Legal lead generation and LegalTech software for motor-vehicle accident, SSDI, VA disability and mass-tort practices.
- Direct a standing team of AI coding agents (Claude Code, Grok CLI, Gemini CLI, Cursor cloud agents). I write the specifications, test what they produce, and approve each change through pull requests.
- MTOS CRM (mtoscrm.com), TypeScript on Node.js: a lead-distribution CRM for mass-tort firms. Each buyer has its own validator; delivery is sent automatically only when a lead meets that buyer's criteria, and duplicates and criteria-not-met leads are tagged rather than lost. It integrates with buyers' Litify instances, captures TrustedForm consent certificates, publishes an OpenAPI specification, and is held to more than 1,000 automated API tests.
- MTOS Portal (mtosportal.com): a vendor portal with tracked referral links, per-vendor attribution, and lead status that updates live from the CRM.
- Case Closed FL: a Next.js personal-injury platform. Built the lead validators for SSDI, Case Closed FL and F4MR traffic.
- tcpa-firewall: a TCPA list-scrubbing service that removes non-compliant records before any outreach begins.
- Launched about twelve legal lead-generation sites, among them caseclosedmva.com (Arizona motor-vehicle accident intake) and caseclosedfl.com. Each uses a multi-step qualifying form, is built for search and AI answer engines, scores about 99 on Lighthouse, and runs on DigitalOcean App Platform.
- Manage the lead-vendor network and LinkedIn outreach to law firms, and apply TCPA, CAN-SPAM and CASL requirements to both.

### ACS Neuro Network Architect, LI MEDIA — Jan 2014 – Present
Online advertising and SaaS.
- Apply business, data and financial analysis to advertising and software engagements.

### Founder, Case Closed FL (https://caseclosedfl.com)
A case-acquisition and attorney-matching service for people injured in motor-vehicle accidents, conceived in Florida and now conducting free eligibility intake across Florida, California, Arizona, Texas and New York.

### Founder, SSDI Campaigns (https://ssdicampaigns.com)
An independent campaign that helps people navigate Social Security Disability Insurance: eligibility, applications, denials and appeals, beginning with a free screening and connecting claimants with qualified representatives.

### IMOS (https://imos-2mi89.ondigitalocean.app)
An institutional platform connecting qualified commercial energy buyers and strategic partners with EN590 diesel sourcing, infrastructure opportunities, and government and corporate relationships.

## AI training and model work

### Mini-Mythos: Domain-adapted coding and reasoning LLM (QLoRA SFT) (Oct 2026 – in progress)
An ongoing supervised fine-tuning run on a 36,000-step schedule, carried out on one RTX 4070 Ti with 12 GB of VRAM.
- Base model: Qwen3-14B (abliterated variant). Base weights are frozen and quantised to 4-bit NF4; low-rank adapters train with bf16 compute using Hugging Face transformers and PEFT.
- At step 10,127, roughly 28% of the schedule, I merged the adapter into the base weights mid-run, converted the checkpoint to GGUF and served it through Ollama, so interim evaluation could proceed while optimisation continues.
- Training data: a ~1.9 GB JSONL instruction/response corpus with a held-out validation split.
- The corpus includes frontier-coding-sft-v1, a 2,640-pair component I curated in-house across eight domains, including advanced coding, software engineering and cybersecurity.
- Retrieval over the same corpus: bge-small-en-v1.5 embeddings indexed in Chroma, top-6 retrieval, exposed through an OpenAI-compatible proxy that supports tool use.
- Supporting work: directed a PyTorch training guide that runs from MNIST to LoRA, QLoRA and GGUF conversion; set up Flux LoRA training in ComfyUI; operate local inference with Ollama, LM Studio, ComfyUI/Flux and Kokoro TTS.

## AI systems and agent orchestration

- Designed and operate a fleet of more than ten coordinated AI agents covering coding, code audit, recruiting, LinkedIn, Microsoft Teams monitoring, Google indexing and media. Scheduled routines assign the work; watchdogs catch what stalls.
- Grok-Memory: a Git-based shared memory and retrieval repository that every agent reads from and writes to, with a companion retrieval project, GENERAL_RAG.
- BOS-Omega, an architecture for orchestrating several LLMs, and AURA-Runtime, an agent runtime written in Rust.

## Additional projects

- LegalHeatmapApp (C#/.NET 8).
- Arabic speech-recognition backend (quran-fastconformer).
- Robotics arm work (SO-ARM100, LeRobot).

## Earlier systems

- Governed LLM gateway with self-auditor: An OpenAI-compatible gateway with provider routing, structured outputs, call logging and a self-auditing stage, later extended into a governed multi-model orchestration platform.
- Evidence-bound agent runtime: A governed runtime in which agents classify intent, plan, select tools and pass risk gates, and in which no task may be marked complete until evidence has been verified and recorded.
- Single-shot AI video engine: A video-generation service for web research, platform-specific copy, image generation and single-shot video prompts, governed by prompt policies, JSON validation, encrypted credentials and audit logs.
- On-device language model: A 4.4-million-parameter BERT model fine-tuned on IMDb, exported to ONNX and quantised to an INT8 model of roughly 16 MB, packaged for on-device Android inference with a Python training pipeline.
- Intake and identity systems: Data-intensive intake and identity workflows spanning accident intake, mass-tort operations and KYC, combining OCR, face matching, tamper checks, schema validation and browser and API testing.
- Schema-driven credit-data parser: A fixed-width Metro 2 parser and renderer driven entirely by schema, accompanied by 126 documented tests.

## Skills

- AI training and evaluation: SFT dataset construction, QLoRA and LoRA fine-tuning, Train/validation splits, Review and approval of AI-generated code, Prompt engineering, Multi-LLM orchestration, RAG (bge-small embeddings, Chroma)
- Languages: TypeScript, Python, C#/.NET, Rust, PowerShell
- Frameworks and ML: Node.js, Next.js, PyTorch, Ollama, LM Studio, ComfyUI/Flux, Kokoro TTS, GGUF conversion
- Engineering practice: OpenAPI specifications, API testing, Pull-request workflow, DigitalOcean App Platform, Namecheap DNS, SEO and Lighthouse performance
- AI tools and APIs: Claude Code, Grok CLI, Gemini CLI, Cursor cloud agents, MCP, Composio, Firecrawl, Exa, Tavily, xAI, Gemini and NVIDIA APIs
- Business and integrations: Zapier, DocuSign, TrustedForm, Litify, Resend, Microsoft Teams, Google Workspace, Business, data and financial analysis

## Blog

Index: https://luislacerda.site/blog/ (RSS: https://luislacerda.site/blog/rss.xml)
- [Consent First: How Luis Lacerda Designs Compliance-Gated Legal Lead Intake](https://luislacerda.site/blog/consent-first-compliance-gated-legal-lead-intake/): A conceptual account of how Luis Lacerda designs legal lead-intake systems in which consent, TCPA compliance and buyer criteria are enforced as gates rather than reviewed after the fact.
- [Directing AI Coding Agents as an Engineering Workforce: Specification, Testing and Review](https://luislacerda.site/blog/directing-ai-coding-agents-as-an-engineering-workforce/): How Luis Lacerda, AI systems architect, directs AI coding agents the way a principal engineer directs a team: written specifications, independent verification and human approval of every change.

## Links

- Website: https://luislacerda.site/
- Résumé: https://luislacerda.site/resume/
- About: https://luislacerda.site/about/
- Ventures: https://luislacerda.site/ventures/
- JSON Resume: https://luislacerda.site/resume.json
- LinkedIn: https://www.linkedin.com/in/luis-lacerda-32a875405/
- Resume PDF: https://luislacerda.site/Luis_Lacerda_Resume.pdf
- Email: intake@abbycrm.com