About Tensorlake Inc
Tensorlake is a serverless platform for agentic applications and Document Ingestion. It provides foundational infrastructure such as serverless compute, code sandboxes, durable execution and unstructured data ingestion APIs to developers in enterprises. We aim to push the frontier of AI that understands real, complex documents and detailed workflows — not just text extraction, but contextual reasoning and scalable application systems.
This role is for an experienced scientist who thrives both in innovating foundational models and building usable, robust AI systems for real users.
- Lead design and experimentation on state-of-the-art models for document understanding, multimodal reasoning, and deep content extraction.
- Research, evaluate, and integrate the latest vision-language models (VLMs), retrieval frameworks, RAG systems, and grounding techniques to drive product impact.
- Develop new benchmarks, datasets, and evaluation methodologies tailored to real-world document AI tasks.
- Build and ship production-ready AI components (e.g., table extraction, structure parsing, semantic indexing, multimodal QA, agentic workflows).
- Work with engineering and product partners to deploy models at scale — from prototype to integrated platform features.
- Collaborate closely with customers and partners to prioritize and validate use cases, including LLM orchestration, context engineering, and agent integration.
Qualifications
Technical Expertise- 5+ years experience in AI/ML research and applied systems; strong record of delivering results.
- Deep background in document understanding, multimodal AI, NLP + computer vision integration.
- Hands-on experience with vision-language models, RAG frameworks, context-aware retrieval, agentic AI, and embedding-based systems.
- Track record of “research → product” delivery: turning prototypes into robust pipelines, APIs, or services.
- Experience optimizing and fine-tuning large models, knowledge of quantization/LoRA/efficient training.
- Proficiency with deep learning frameworks (PyTorch preferred), Python, and scalable ML tooling.
- Experience with open-source frameworks and community contributions
- Background in agents, RAG architectures, retrieval systems, and context engineering.
- Experience designing benchmarks, quality metrics, or curated datasets for complex tasks.
What Sets You Apart
- You balance deep technical curiosity with product focus and can speak fluently to both ML research and engineering issues.
- You’ve shipped models and systems that are used by developers or customers in real scenarios (not just research demos).
- You are comfortable working in a fast-moving startup environment where priorities evolve and innovation is part of the culture.
Benefits
- - Ability to save in 401(k) plans
- - Comprehensive Healthcare and Dental Benefits
We’re building something foundational for how companies integrate AI into knowledge workflows, RAG pipelines, document agents, and real-world enterprise applications. If you’re excited about shaping AI products that matter, and you enjoy bridging research with engineering execution, this is the role to join. LinkedIn
Skills Required
- 5+ years experience in AI/ML research and applied systems
- Deep background in document understanding, multimodal AI, and NLP + computer vision integration
- Hands-on experience with vision-language models, RAG frameworks, context-aware retrieval, agentic AI, and embedding-based systems
- Track record of turning prototypes into robust pipelines, APIs, or services
- Experience optimizing and fine-tuning large models; knowledge of quantization, LoRA, and efficient training
- Proficiency with Python and scalable ML tooling
- Proficiency with deep learning frameworks (PyTorch preferred)
- Experience designing benchmarks, quality metrics, or curated datasets for complex tasks
- Open-source framework contributions or community involvement
What We Do
Tensorlake is a platform for building Agentic automation and applications in Enterprises. We provide three foundational primitives to build reliable Agents: serverless compute with durable execution, code sandboxes and accurate document ingestion for applications. Agentic Runtime: Build agents with any framework, and deploy them on our serverless platform to expose them as an HTTP API. The runtime includes durable execution to replay requests and resume execution from where they crashed. The runtime offers code sandboxes to securely run LLM generated code in agents. Document AI: VLM-powered extraction that understands document semantics. Handles multi-page tables, handwriting, nested layouts, and strikethrough text. Returns layout-aware Markdown or validated JSON. Enterprise-Ready: HIPAA | SOC 2 Type II | Used by financial services, healthcare, insurance, logistics and legal tech where accuracy and reliability are non-negotiable. No Messy Infrastructure Tax: No Airflow. No Spark. No queue orchestration. No container management. Just Python that runs durably at global scale.
.jpg)
