About Deeli AI
In an era of data explosion and AI breakthroughs, innovation decisions are still made based on fragmented signals and gut instinct. Every year, global R&D focused enterprises spend billions exploring new technologiesyet they often miss key opportunities or bet on the wrong ones.
This isn't just an efficiency problem. It's a massive waste of innovation potential.
At Deeli AI, we're solving this challenge by building the world's first AI-native innovation intelligence platform. Our mission is to empower every innovation-driven organization to make confident, data-backed decisionsat the right time, with the right signals. We aggregate global deep tech signalsacademic research, patents, grants, and stealth startupsto surface emerging technologies before the market reacts.
Founded by a second-time entrepreneur and built by a world-class team with experience at MediaTek, AWS, and Deloitte, we're backed by the Berkeley SkyDeck Fund in Silicon Valley, ITIC (a deep tech fund in Taiwan), and NVIDIA Inception. We're already collaborating with multiple top semiconductor companies worldwide, with pilots valued in the hundreds of thousands of dollars.
This is the perfect time to join usat an inflection point where your contribution will directly shape both our product and the future of innovation intelligence.
About the Role
We're looking for a full-time Data Scientist to join us in refining the intelligence core of our platform. You'll work closely with our dev and product teams to design the logic, scoring systems, and algorithms that power our search and recommendation engine.
This is a hands-on, high-impact role. You will be trusted to architect the ranking logic. You will start by using smart heuristics, business rules, and LLMs as reasoning engines to solve immediate challenges like the Cold Start problem. As we scale, you will help us optimize these systems for cost and speed. If you enjoy solving messy data problems with pragmatic solutions and shaping product direction in a fast-moving environment, we'd love to meet you.
Your Responsibilities
- Collaborate with engineering and product teams to design and deploy the ranking logic for our analysis platform.
- Design and implement LLM-as-a-Judge pipelines, using LLMs (via API) with strict rubrics to score content for technical depth and accuracy, using these scores to directly influence search ranking.
- Develop deterministic algorithms and heuristics (e.g., freshness boosters, weighted scoring) to solve the Cold Start problem for new content.
- Build and maintain the Python/SQL data pipelines required to extract features that influence search rankings.
- Establish evaluation metrics and Gold Standard datasets to scientifically measure improvements in search accuracy and relevance.
- Eventually explore traditional ML methods (scikit-learn) or smaller models to distill our expensive LLM pipelines into faster, cheaper solutions.
- If interested, optionally support other areas such as NLP entity extraction, RAG (Retrieval Augmented Generation), or vector search optimization.
Must-Have Qualifications
- 2+ years of experience in Data Science, Engineering, or a relevant technical field.
- LLM & Prompt Engineering Proficiency: Strong experience using LLM APIs (OpenAI, Anthropic) programmatically. You know how to design prompts to get structured, reliable outputs (e.g., JSON) and use them as components in a larger algorithm.
- Strong Python Foundation: Fluent in data manipulation (Pandas, NumPy) and comfortable writing clean, production-level code (not just Jupyter notebooks).
- Logical & Mathematical Intuition: Ability to design scoring rules or weighted algorithms (e.g., _'How do we mathematically combine an LLM quality score with a recency score'_).
- Data Fluency: Proficient in SQL and capable of handling complex datasets.
- Language: Fluent in English and Mandarin, both spoken and written, able to communicate effectively in global teams.
- Mindset: Pragmatic. You value reliable results over complex architectures.
- Search Domain Knowledge: Understanding of Information Retrieval concepts (TF-IDF, BM25) and why search engines fail.
Preferred Qualifications
- Traditional ML Knowledge: Familiarity with scikit-learn and statistical modeling. (We view this as an optimization tool for the future).
- MLOps Experience: Familiarity with tracking experiments (MLflow) or managing model deployment loops.
- Vector Tech: Familiarity with Vector Databases (pgvector, Qdrant) or Search Engines (Elasticsearch).
- Experience with cloud data infrastructure (GCP BigQuery, Data Lakes) or basic containerization (Docker).
Compensation
We offer a highly competitive compensation package, consisting of base salary and stock options, along with comprehensive insurance coverage. For the right candidate, we're open to tailoring the structure to reflect your experience, ownership mindset, and potential impact on the company.We believe in rewarding talent not just fairlybut boldly. You'll be compensated like a key contributor, not a cog in the machine.
Why Join Deeli AI
- Shape the intelligence layer: Design the ranking logic, scoring systems, and algorithms that directly power how users discover breakthrough technologies
- High ownership, real impact: Your heuristics and LLM pipelines won't just be experimentsthey'll ship to production and define our search quality from day one
- Work at the frontier of applied AI: Use LLMs as reasoning engines, build evaluation frameworks, and eventually optimize into scalable ML solutionsall on real, messy innovation data
- Collaborate across the stack: Partner closely with engineering and product to turn your data science work into features that solve real problems for enterprise R&D teams
- Global momentum, early stage: Join a well-funded, fast-growing team already working with leading corporate R&D teams in the U.S., Europe, and Asia
