Staff ML Engineer

About Discovery & Scout

The Discovery team sits at the core of Harmonic’s product and data engine. We make it possible to search, research, analyze, compare, and evaluate across the entire universe of startups (we track tens of millions of companies, hundreds of millions of operators and, tens of thousands of investors along with their deal activity, and counting). We source and ingest data from first-, second- and third-party sources, including hundreds of fragmented and unstructured sources, and turn them into structured insights and searchable indices. Recently, we built the industry leading agentic layer for startup search and research, that’s unlocking new paradigms for VCs to make sense of this vast amount of data on startups and automate critical tasks in their sourcing and diligence workflows.

We’re in the middle of a massive platform shift where AI and LLMs are transforming what "software" is. Agentic product- and user-experience patterns will be spelled out in playbooks before long – and we’re building those playbooks, not waiting for them. This includes conversational interfaces, deep research, dynamic agentic frameworks purpose built for venture, and AI-first product workflows. We’re looking for folks that want to help redefine software and push on this frontier with deep curiosity, an insatiable growth mindset, and high agency.

To learn more about the team:

• Explore Working with Sang
• Check out your Scout focused teammates Austin, Yesh, and Kelvin
• Explore the Team Page. We’re a mix of ex-founders and seasoned engineers from top engineering institutions like Google, LinkedIn, Microsoft and Meta

The role

Scope of the role:

• Own semantic understanding: Semantic search is one of our most valuable features - it enables our users to describe a company or market problem as they see fit, and find the most compelling startups in that space. It still has so much room for improvement, and every margin of improvement translates to our customers finding better opportunities, and more innovation getting funded.
  • You will lead the strategy and roadmap to improve and evolve semantic search precision and recall outcomes. This spans everything from structuring data for embeddings and fine-tuning embedding models through to creative retrieval and ranking strategies. You'll experiment with and implement novel approaches to continually increment towards perfect search results.
• Architect RAG + personalization: Harmonic is becoming a centralized repository for all of a customer’s startup engagement and investments. As we contextualize user’s implicit interactions and explicitly provided content, we can further personalize search results and research answers, helping them find the most relevant opportunities that align with their investment thesis.
  • To drive this, you will own designing and building systems that persist and leverage high value data to produce outcomes with real-time, individualized responses. Think storage and RAG over data like: notes, pitch decks, correspondences – plus prior search queries, engagement patterns, and long-term-memory stores.
• Cross-team leadership: Partner with infra to generalize your search and RAG architecture for platform-wide use, and drive integration in features across product engineering teams.

Background we’re looking for:

• 5+ years of experience as a machine learning engineer, with flexibility for exceptional candidates
• Proven track record optimizing semantic search, embeddings, vector stores, or ranking pipelines to drive measurable improvements in relevance and recall
• Hands-on experience with LLM-powered search, RAG implementations, or agentic workflows
• Strong systems-level thinking: you’ve designed data models, chosen indexing strategies, and built serving paths optimized for both performance and user experience

Pay

$210k-$280k Salary + Equity depending on the level

Our stack

The Process

Here’s our interview process:

1. Recruiter Screening: 20 mins
2. Technical Screening (Debugging): 45 mins
3. Technical Screening (System Design): 45 mins
4. Jam Session: 2 hours 
5. Behavioral interviews with our Head of Engineering and CEO: 30 mins each
6. References