Founding AI/ML Developer

OpenSourceAGI (domain https://opensourceagi.app/) is hiring an AI/ML Developer to build open, citation-grounded research agents that turn the web, documents, and collective knowledge into useful, inspectable intelligence. This is an early-stage, equity-only founding-role opportunity, with up to 5–10% equity based on experience, commitment, scope of responsibility, and founder fit. Following successful fundraising, the intended cash compensation is $150,000/year, subject to funding, board/company approval, business conditions, and the final employment agreement.

OpenSourceAGI is an open-source research and intelligence platform reimagining the internet as a self-organizing mind map. Our products help people search across the web and academic sources, extract and cite material from articles, PDFs, and videos, organize research in collaborative editors, and use AI agents to develop transparent research outlines and answers.

The platform is available at opensourceagi.app. We are actively preparing and applying to venture-capital investors and startup funding programs. The person in this role will have an opportunity to help shape the product, technical direction, fundraising narrative, and early company foundation.

Compensation note: Until OpenSourceAGI closes an equity-financing round or otherwise secures sufficient funding, this role is compensated in equity only—there is no salary, hourly wage, or guaranteed cash compensation during the pre-funding period. The anticipated post-funding salary target is $150,000 annually, but it is not guaranteed and depends on successful fundraising and a formal employment agreement.

Role structure
Item Details
Title Founding AI/ML Developer / Founding AI Research Engineer
Stage Pre-funding, early-stage startup
Initial compensation Equity only
Equity range Up to 5–10% equity, based on role scope, seniority, expected commitment, vesting, and final agreement
Salary after funding Target compensation of $150,000/year after successful funding, subject to company resources and formal agreement
Funding activity OpenSourceAGI is actively preparing and applying to VC firms, accelerators, grants, and other startup funding opportunities
Working arrangement Flexible; scope, hours, location, and level of commitment to be agreed before engagement
Employment status This posting does not itself create employment, compensation, or a binding offer; final terms require written agreements
What you’ll build
You will help define and ship the intelligence layer behind OpenSourceAGI, including:

Research agents that search, read, rank, extract, summarize, cite, and synthesize information from the open web.

Retrieval and reranking pipelines that select high-quality sources, passages, quotes, and evidence for a user’s research question.

Agent workflows using multiple LLM providers and open-weight models, with routing based on capability, latency, cost, privacy, and task type.

Citation-grounded answer generation that connects claims to the exact source passages used to support them.

Document-understanding systems for HTML, PDFs, DOCX, Google Docs, YouTube transcripts, and other uploaded or linked sources.

Evaluation frameworks for research quality: factuality, source quality, citation precision and recall, completeness, latency, cost, and user trust.

Semantic search, embeddings, topic models, clustering, entity extraction, and knowledge-graph features that power OpenSourceAGI’s self-organizing research maps.

AI features for the REASON writing environment, including quote recommendations, outline generation, evidence discovery, writing assistance, and research-memory workflows.

Privacy-conscious personal knowledge-base systems that let users organize the information, sources, and perspectives that influence their thinking.

Open-source packages, APIs, MCP tools, and developer documentation that allow other developers and AI agents to build on OpenSourceAGI.

What you’ll do
Design, implement, test, and operate production AI systems from prototype through deployment.

Build robust RAG and deep-research pipelines that combine web search, content extraction, ranking, structured planning, tool use, memory, and answer generation.

Develop methods for extracting clean, structured content and metadata from difficult real-world sources, including scanned or complex PDFs, JavaScript-heavy pages, long articles, citations, tables, and video transcripts.

Improve source ranking using retrieval models, heuristic quality signals, domain reputation data, metadata, freshness, user intent, and learned rerankers.

Build reliable LLM orchestration that supports providers such as Anthropic, OpenAI, Google, Groq, xAI, local Llama-family models, and other compatible APIs.

Create model-routing policies that choose the right model and workflow for a task while maintaining quality, controlling cost, and preserving user choice.

Develop offline and online evaluation datasets, test suites, and dashboards for agent behavior.

Measure hallucination, unsupported claims, citation correctness, answer usefulness, extraction accuracy, tool-call reliability, and regression risk.

Help establish clear standards for transparent AI behavior: distinguish sourced facts from inference, expose provenance, communicate uncertainty, and avoid presenting unsupported claims as evidence.

Build embedding, vector-search, hybrid-search, clustering, topic-modeling, and graph-based systems for long-running research collections.

Collaborate with product and full-stack engineers to turn ML capabilities into fast, understandable user experiences.

Contribute production-quality code to a TypeScript-first monorepo and publish reusable open-source packages where appropriate.

Write technical documentation, reproducible benchmarks, and clear implementation notes for contributors and users.

Participate in architecture decisions across Cloudflare Workers, serverless APIs, queues, durable state, object storage, databases, browser rendering, and optional GPU/container workloads.

Core technical areas
Area Examples of work
LLM systems Tool calling, agent planning, structured outputs, streaming, prompt engineering, model routing, guardrails, context management
Retrieval Embeddings, BM25/hybrid search, query rewriting, rerankers, chunking, metadata filtering, passage selection
Grounded generation Inline citations, quote extraction, evidence attribution, claim verification, uncertainty handling
Document AI HTML extraction, PDF parsing, OCR workflows, document structure recovery, citation metadata extraction
Search intelligence Multi-engine search, result deduplication, domain quality scoring, freshness ranking, query expansion
Knowledge systems Topic modeling, semantic clustering, entity resolution, research graphs, personalized knowledge bases
Evaluation Golden datasets, LLM-as-judge with human calibration, regression tests, trace analysis, cost and latency measurement
Production ML API design, observability, caching, rate limits, queues, background jobs, model-serving integration
Developer tools MCP servers, SDKs, OpenAPI clients, CLI tools, reusable packages, example applications
Required qualifications
Strong software-engineering experience building and shipping production systems.

Professional or substantial project experience with LLM applications, NLP, information retrieval, search, RAG, document intelligence, or agentic workflows.

Strong TypeScript or JavaScript skills, with the ability to work effectively in a modern monorepo.

Practical Python proficiency for ML experiments, data pipelines, evaluation, model tooling, or GPU-backed services.

Experience integrating LLM APIs and/or running open-weight language models locally or in hosted infrastructure.

Familiarity with embeddings, vector databases or vector indexes, retrieval pipelines, reranking, and chunking strategies.

Experience designing systems that process messy, unstructured, real-world content.

Strong judgment about reliability: you test assumptions, measure quality, instrument systems, and care about failure modes.

Ability to communicate complex AI-system behavior clearly to engineers, contributors, and nontechnical users.

Comfort working in a fast-moving startup and open-source environment where requirements may evolve as users reveal what is valuable.

A portfolio, GitHub profile, shipped product, research project, open-source contribution, or technical write-up demonstrating relevant work.

Preferred qualifications
Experience building web-search, deep-research, or citation-answering systems.

Experience with retrieval evaluation, ranking metrics, query relevance, or search-quality analysis.

Experience with modern agent frameworks and protocols such as the Vercel AI SDK, Mastra, MCP, LangChain, LangGraph, or equivalent custom orchestration systems.

Experience deploying on Cloudflare Workers, Durable Objects, Queues, R2, D1, KV, Browser Rendering, or comparable edge/serverless platforms.

Familiarity with Hono, Next.js, React, Bun, Tauri, Docker, and API-first development.

Experience with PDF extraction, OCR, Docling, PDF.js, browser automation, web scraping, or content-cleaning pipelines.

Experience with open-source NLP and ML tooling such as Hugging Face Transformers, sentence-transformers, PyTorch, vLLM, Ollama, llama.cpp, or similar.

Familiarity with graph algorithms, entity linking, topic modeling, community detection, or visualization of large information collections.

Experience working with datasets involving argumentation, claim-evidence relationships, debate, research synthesis, or scientific literature.

Experience with safety, trust, provenance, privacy, and misuse considerations for AI research tools.

Contributions to open-source packages, SDKs, APIs, developer tools, or technical communities.

Your first 90 days
First 30 days
Become productive in the OpenSourceAGI codebase, local development workflow, API surface, and existing agent packages.

Map the current research pipeline from search query through extraction, source selection, model calls, citations, and final response.

Establish or improve baseline evaluation cases for web research, source extraction, citation correctness, and long-context synthesis.

Ship an initial improvement to a high-impact quality issue such as passage ranking, citation formatting, extraction robustness, latency, or agent observability.

Days 31–60
Own a major AI capability or quality track, such as hybrid retrieval and reranking, research-agent planning, citation verification, PDF/document intelligence, or topic-map generation.

Build a measurable evaluation loop with representative datasets and regression protection.

Improve the production agent pipeline with concrete gains in answer quality, grounding, speed, reliability, or unit cost.

Work closely with the product team to expose new AI functionality through an intuitive workflow rather than a “black box” chat experience.

Days 61–90
Launch a substantial feature or infrastructure improvement that users can depend on.

Document the architecture, quality metrics, limitations, and operational playbook.

Help define the next-quarter AI roadmap based on measured user needs, product behavior, and technical leverage.

Contribute reusable components, examples, or documentation that strengthen the OpenSourceAGI open-source ecosystem.

How we work
We value evidence over hype. A feature is not “intelligent” merely because it calls a large model; it should be measurable, inspectable, useful, and honest about uncertainty.

We build in the open where practical. Clear documentation, reproducible work, thoughtful issues, and high-quality pull requests are part of the product.

We favor pragmatic systems. Use a frontier API when it provides leverage, an open model when control or cost matters, and deterministic methods when they are more reliable.

We optimize for researcher trust. Sources, quotes, provenance, and transparent reasoning matter as much as fluent answers.

We expect ownership. You will identify problems, propose solutions, make tradeoffs explicit, ship improvements, and learn from production behavior.

We care about user agency and privacy. Users should have meaningful control over their data, model choices, memories, and research workflows.

Tech environment
OpenSourceAGI is built around a modern, AI-native web stack:

TypeScript, JavaScript, Bun, Node.js, and Python

Next.js, React, shadcn/ui, Tauri, and browser extensions

Cloudflare Workers, D1, R2, KV, Durable Objects, Queues, Browser Rendering, and edge/serverless deployment patterns

Hono, OpenAPI, typed API clients, MCP servers, and package-based architecture

Multi-provider LLM support through Vercel AI SDK–style integrations

Web search, browser rendering, article extraction, PDF-to-HTML conversion, YouTube transcript extraction, and document metadata/citation pipelines

SQLite/Turso-style persistent storage and collaborative-editor workflows

GitHub Actions, automated testing, coverage, package publishing, and open-source collaboration

What success looks like
Success means researchers receive better answers because the system finds stronger sources, extracts more accurate evidence, cites it precisely, explains its confidence, and makes the reasoning workflow easier to inspect and improve. It also means developers can build on OpenSourceAGI through reliable APIs, packages, MCP tools, and well-documented AI primitives.

You will help make AI-assisted research less opaque, more useful, and more accountable—while building infrastructure that can support a growing ecosystem of research tools, personal knowledge systems, and open-source agents.

Application
Please share:

Your résumé, LinkedIn, GitHub, portfolio, or equivalent body of work.

A brief note on a production AI/ML system, search/retrieval project, agent, or open-source contribution you are proud of.

One example of how you would evaluate whether a citation-grounded research agent is actually improving.

Your preferred working arrangement, location/time zone, availability, and compensation expectations.

We welcome applicants with nontraditional backgrounds. If you have built strong systems, shipped useful tools, conducted meaningful research, or contributed to open source, we encourage you to apply.