μμ μ΄ν΄ AIμ κΈλ‘λ² κΈ°μ€μ ν¨κ» λ§λ€μ΄ κ° μΈμ¬λ₯Ό μ°Ύμ΅λλ€!
νΈμ°λΈλ©μ€λ λ°©λν μμ λ°μ΄ν°λ₯Ό ν¨κ³Όμ μΌλ‘ μ²λ¦¬νμ¬, μμμ νΉνλ κ²μ, λΆμ, μμ½, μΈμ¬μ΄νΈ μμ± κΈ°λ₯μ μ 곡νλ μΈκ³ μ΅κ³ μμ€μ μμ νΉν AI λͺ¨λΈμ λ§λ€κ³ μμ΅λλ€.
μΈκ³ μ΅λ μ€ν¬μΈ 리그μμλ νΈμ°λΈλ©μ€ λͺ¨λΈμ νμ©ν΄ λ°©λν κ²½κΈ° μμ μμμ λΉ λ₯΄κ³ μ ννκ² νμ΄λΌμ΄νΈλ₯Ό μ λ³νμ¬ μ΄κ°μΈνλ μμ² κ²½νμ μ 곡νκ³ μμ΅λλ€. κ΅λ΄ ν΅ν©κ΄μ μΌν°μμλ μκΈ° μν©μ μ μν λμνκΈ° μν΄ νΈμ°λΈλ©μ€μ ν¨κ» CCTV μμμ ν¨μ¨μ μΌλ‘ νμνκ³ μμΌλ©°, μ μΈκ³ μ£Όμ λ°©μ‘μ¬μ μ€νλμ€λ€μ μμμ΅ λͺ μ μμ²μλ₯Ό μν μ½ν μΈ μ μμ νΈμ°λΈλ©μ€ λͺ¨λΈμ νμ©νκ³ μμ΅λλ€.
νΈμ°λΈλ©μ€λ μνλμμ€μ½μ μμΈμ μ€νΌμ€λ₯Ό λ Deep Tech μ€ννΈμ μΌλ‘, 4λ μ°μ CB Insights μ μ μΈκ³ 100λ AI μ€ννΈμ μ μ΄λ¦μ μ¬λ Έμ΅λλ€. NVIDIA, NEA, Index Ventures, Databricks, Snowflake λ± μΈκ³μ μΈ VCμ κΈ°μ λ€λ‘λΆν° μ΄ 1μ΅ 1μ²λ§ λ¬λ¬ μ΄μμ ν¬μλ₯Ό μ μΉνμΌλ©°, νκ΅μμ κ°λ°λ AI λͺ¨λΈ μ€ μ μΌνκ² Amazon Bedrockμ ν΅ν΄ μλΉμ€λ©λλ€. μ°λ¦¬λ νμν λλ£λ€κ³Ό νμ μ μΈ μ νμ λ§λ€κ³ μ μΈκ³ κ³ κ°λ€κ³Ό ν¨κ» μ±μ₯νκ³ μμ΅λλ€.
νΈμ°λΈλ©μ€λ λ€μκ³Ό κ°μ ν΅μ¬ κ°μΉλ₯Ό μ€μ¬μΌλ‘ μΌν©λλ€.
λμ νμ λν΄ μ μ§νκ³ μ±μ°°ν μ μλ νλ
μ€ν¨μ νΌλλ°±μ λλ €μνμ§ μλ λκΈ°μ κ²Έμ
λμμλ νμ΅μ ν΅ν΄ νμ μλμ ν¨κ» λμ¬ κ°λ μμΈ
λμ μ μΈ λ¬Έμ λ₯Ό ν¨κ» ν΄κ²°νλ©° μ±μ₯νλ κ³Όμ μ μ¦κΈ°λ λΆμ΄λΌλ©΄, κ·Έ κΈ°νκ° μ¬κΈ° νΈμ°λΈλ©μ€μ μμ΅λλ€.
νΈμ°λΈλ©μ€μ λ©ν°λͺ¨λ¬ νν νμ΅(Representation Learning)κ³Ό νλ‘λμ μλΉμ λ΄λΉνλ νμ λλ€. λΉλμ€, μ€λμ€, ν μ€νΈ λ± λ€μν λͺ¨λ¬λ¦¬ν°λ₯Ό νλμ μλ² λ© κ³΅κ°(Embedding Space)μ ν΅ν©νλ λͺ¨λΈμ νμ΅νκ³ , μ΄λ₯Ό μ μΈκ³ μμ² κ³ κ°μ΄ μ¬μ©νλ νλ‘λμ μμ€ν μΌλ‘ μμ μ μΌλ‘ μλΉν©λλ€.
λκ·λͺ¨ λΆμ° νμ΅ νκ²½μμ λ©ν°λͺ¨λ¬ μλ² λ© λͺ¨λΈμ μ€νμ μννκ³ , μ°κ΅¬ κ²°κ³Όλ₯Ό μ€μκ° μΆλ‘ μμ€ν μΌλ‘ μ ννλ End-to-End κ³Όμ μ μ± μμ§λλ€. NVIDIA B300 λ± μΈκ³ μ΅κ³ μμ€μ GPU 리μμ€μ λν μ κ·Ό κΆνμ λ°νμΌλ‘, μ°κ΅¬μμ νλ‘λμ κΉμ§μ μ ν μ£ΌκΈ°λ₯Ό μ΅μνν©λλ€.
μ°κ΅¬ κ²°κ³Όκ° μκ°μ λ΄μ μ μΈκ³ κ³ κ°μκ² μ 곡λλ μ§§μ κ°λ° μ¬μ΄ν΄ μμμ, Research, Product, Infrastructure νκ³Ό κΈ΄λ°ν νμ νλ©° κΈ°μ μ μν©νΈλ₯Ό λ§λ€μ΄κ°λλ€.
As a Senior MLE on the Embedding & Search team, you will own and build key components of TwelvaLabs' search and retrieval platform β the systems that combine vector search, lexical retrieval, and reranking into fast, accurate, and scalable search experiences for our customers.
This is a systems-heavy ML engineering role at the intersection of information retrieval, ML serving, and distributed systems. We're looking for a strong engineer who can take well-scoped problems with moderate ambiguity, break them down into concrete milestones, and deliver reliable, performant solutions.
Own and build core subsystems of our search platform on EKS β spanning vector indexing (ANN), lexical retrieval, hybrid fusion, reranking, and temporal (segment-level) search
Optimize retrieval performance at million to billion-scale across both vector and lexical paths
Develop and maintain production microservices across the search stack
Collaborate with the research/training team to co-evolve embeddings, reranking models, and retrieval strategies
Implement and maintain evaluation frameworks for search quality (recall, precision, latency, relevance)
Work cross-functionally with platform/infra and product teams to ship search capabilities end-to-end
6β8 years building production ML systems, with emphasis on search, retrieval, or recommendation
Strong software engineering skills in Python; Go experience is a plus
Hands-on experience with ML model serving and inference optimization in production (e.g., KServe, Triton, Ray Serve)
Experience with information retrieval systems β embedding-based search, lexical search (BM25/Elasticsearch), or hybrid retrieval
Proficiency with data pipelining and orchestration (Spark, Ray, Airflow, Kubeflow, or similar)
Strong Kubernetes experience and familiarity with databases, vector databases, and search engines
Solid distributed systems and async programming fundamentals
Good English communication skills (verbal and written)
Experience with multimodal or video search/retrieval systems
Familiarity with temporal indexing or segment-level retrieval (shot boundary detection, scene search)
Experience with hybrid retrieval strategies (rank fusion, reranking models, score normalization)
Experience with ANN index tuning at scale
Experience building services with high-demand SLAs
twelve-labs