Open Doors: Looking for Independent AI/ML Engineers and Open-Source Contributors

LEX AI is opening its platform as open source. We welcome strong engineers — AI/ML, backend, data, frontend — to contribute or join the team. What's already open, who we're looking for, and how to get involved.

Open Doors: Looking for Independent AI/ML Engineers and Open-Source Contributors

LEX AI has been built since 2024 by a small team. We're now opening part of the platform as open source and inviting independent engineers to join — as contributors and as future team members.

What LEX AI Is

LEX is a Ukrainian legal AI platform. Semantic search across 100M+ court decisions (EDRSR — the largest open court decisions corpus in Europe), legislation from the Ukrainian Parliament, OSINT and due diligence, consultations, billing. The stack is assembled as MCP (Model Context Protocol) servers behind a unified gateway.

Our second product — Panoptic (panoptic.com.ua) — is an OSINT platform aggregating 18+ intelligence data sources: sanctions, corporate ownership, credential breaches, IP/domain reputation, GDELT, INTERPOL, World Bank Debarment.

We're building Harvey.ai-level quality for Ukrainian jurisprudence on open-weight models — DeepSeek-V3, Llama, Qwen — because the data is unique (no such corpus exists in the EU), and open-weight models after continued pre-training deliver 90%+ of flagship LLM quality on domain tasks at a fraction of the cost.

Our Repository Layout

We maintain two repositories, and this is important to understand up front.

1. overthelex/secondlayer — public, open source

The main monorepo, now public:

https://github.com/overthelex/secondlayer

Almost the entire platform is there:

• Three MCP servers (mcp_backend, mcp_rada, mcp_openreyestr) — court cases, parliament, business registry
• Web frontend (lexwebapp) — React 19, Vite, TailwindCSS, Zustand, TanStack Query
• Shared TypeScript package (packages/shared) — LLM manager, logger, cost tracker, SSE handler, database base class
• Developer Console (platform) — platform.legal.org.ua, the developer portal: API keys, documentation, integration examples
• Data importers for 340M+ records from 15 government APIs — EDRSR, Verkhovna Rada, NACP, OpenReyestr, OpenSanctions, GLEIF, ICIJ Offshore Leaks, HIBP, NVD, INTERPOL, World Bank
• Full CI/CD — self-hosted GitHub Actions runner, blue-green deploy over SSH, Claude Code auto-fix agents for failing builds
• All deployment configuration — Docker Compose for local, blue-green compose for production, nginx, manage-gateway script
• Playwright E2E + Jest/Vitest unit tests
• Migrations for three PostgreSQL instances
• Internal documentation, architecture notes

Clone it, read it, run it locally. Everything needed for a working instance is there.

2. overthelex/secondlayer-core — private, closed source

A separate repository we deliberately keep private. It contains:

• Chat and orchestration logic — how user queries are classified, routed between tools, and composed into multi-step responses
• Production prompts — exact templates, few-shot examples, system messages used in production for classification, summarization, citation checks, tool selection
• Billing and payment business logic — credit deduction rules, subscription tier resolution, Monobank callback handlers
• Anti-abuse and rate-limiting heuristics we don't want adversaries to enumerate

This is the minimum closed surface that protects our product positioning without holding back the open parts. The whole "chat logic" — prompt engineering, tool orchestration, model cascading, response composition — lives here, and it is not public. The open repository expects this layer as a dependency but ships fully functional stub implementations for contributors.

If you join the team, you get access to secondlayer-core from day one. If you contribute externally, you work against the open repo and the stubs — that already covers everything except production prompt engineering.

Who We're Looking For

We don't hire by job title. We're looking for people who already do strong work — and want to do it on a meaningful domain, with real data and real users.

AI/ML engineers:

• LoRA fine-tuning of large models (70B+), continued pre-training
• Embeddings fine-tuning (BGE-M3, custom encoders) for retrieval
• RLHF, constitutional alignment, adversarial training setups
• Hands-on with Vertex AI / SageMaker HyperPod / Trainium / TPU v5p on multi-node clusters
• Retrieval-augmented generation, citation verification, hallucination guards

Backend / distributed systems:

• PostgreSQL at billion-row scale (pgvector, partitioning, TOAST optimizations)
• Event-driven architectures, queues, replication, PgBouncer
• MCP servers, tool orchestration, LLM gateways, cost tracking

Data engineering / OSINT:

• Scraping at scale (rate-limiting, proxy rotation, resume logic, checkpointing)
• ETL for government open registries
• Sanctions screening, KYC/AML, due diligence pipelines

Frontend:

• React 19 + TypeScript at production level
• Complex UI for legal analytics (data-heavy dashboards, evidence panels)
• Ukrainian i18n, accessibility, performance optimization

Philosophy

• Open everything that doesn't break the business. We don't hide the architecture — it isn't the competitive edge. The edge is data, domain quality, and iteration speed.
• Pragmatism over hype. A distributed monolith today can be the right answer. Microservices ≠ virtue. A framework ≠ a solution.
• Legal deserves serious AI engineering. Not "a chatbot with statutes" — real legal modeling: constitutional alignment, citation verification, jurisdictional specialization.
• Open source by default. If the code doesn't contain proprietary prompts, API keys, or client data — it's public.

How to Join

As a contributor:

1. Check open issues on GitHub (github.com/overthelex/secondlayer)
2. Submit a PR — we review within 48 hours
3. For large changes, open a discussion first

As a hiring candidate:

Email vladimir@legal.org.ua with a short resume. No page-long cover letter needed — show three things:

1. What you've done before (GitHub, a link to a specific project with detail)
2. Why this domain — legal AI, open data, OSINT — interests you
3. What you want to build in the next 6 months

We respond fast. Interview is a technical discussion (no LeetCode), a pair-programming session on a real task from the backlog, and a coffee chat with the team.

Our Promise

• Fully remote. The team is distributed across Europe.
• No micromanagement. Trust by default. Output matters more than Slack presence.
• Prod access from day one. No "probation month" in read-only.
• Compute budget. If an idea needs a GPU cluster — we talk to Google Cloud, AWS, Nebius and find the resource.
• Publication under your name. Your work is your credit. We don't hide contributors.

Context

We're currently in active conversations with Google Cloud and AWS about sponsorship for a 12-month ML training plan (195K–265K, DeepSeek-V3 685B continued pre-training on 50–80B tokens of the EDRSR corpus). We have paying users and B2B clients. Not a startup-in-a-garage, not another enterprise clone. Something in between — and that's what makes the work interesting.

If you're excited by building real AI infrastructure for jurisprudence on the largest open court decisions corpus in Europe — let's talk.

Open repo: https://github.com/overthelex/secondlayer Closed core (chat logic): overthelex/secondlayer-core — private, granted on hire Contact: vladimir@legal.org.ua Site: https://legal.org.ua