Led · Equity · Jan 2024

42Maru is a generative AI startup focused on mitigating large language model hallucinations through its RAG42 (Retrieval-Augmented Generation) and MRC42 (machine reading comprehension) engineering. The company offers enterprise-friendly features such as a Private mode to protect internal and sensitive customer data. It also builds lightweight, industry-specialized models to slash the costs of solution development, training, and serving. 42Maru provides end-to-end services from RAG consulting to LLM solution implementation, targeting domain-specific engine development. The firm says its technical strength and multiple proven commercial use cases were key to attracting investment. 42Maru plans to deepen partnerships to accelerate commercialization and expand industry-focused deployments. 42Maru develops artificial intelligence–based Deep Semantic question-and-answer (Q&A) solutions that produce a single, contextually understood answer rather than multiple candidate answers. Its product suite includes a deep learning–based machine reading solution (MRC42) and a search-augmented creation solution (RAG42). The company focuses on upgrading AI solutions for business customers and expanding B2B offerings. 42Maru will engage in joint research and development on large language models and AI-related technologies with partners. The partnership aims to combine 42Maru’s machine reading and RAG solutions with communication-specific language models to create integrated services. The article reports a strategic equity investment and business cooperation to support these plans.

Participated · Series B · Dec 2023

Allganize provides enterprise LLM solutions including the Alli LLM App Market, the Alli AI chatbot, and Alli for Invoice to help companies automate and optimize workflows. The Alli LLM App Market offers 100+ ready-to-use LLM apps and no-code app creation, with multi-LLM support (GPT-4, GPT-3.5, Llama2, PaLM2) and options for sLLMs and on-premise deployment. The Alli chatbot delivers high response accuracy, document-based QA extraction, external tool integrations, and is offered from ¥100,000 per month. Alli for Invoice extracts data from non-standard invoices without complex setup and supports required invoice items for Japan’s invoicing system. Allganize plans to further advance its AI solutions, accelerate global expansion focused on Japan, the U.S., and Korea, and aggressively hire AI talent. The company was founded in 2017 and operates from Tokyo, Houston, and Seoul; it recently raised $20.2M in a Series B to support product development and international growth. Allganize develops NLP technology that extracts answers to natural-language questions directly from millions of unstructured text documents without manual preparation of training data. The platform can also automatically generate questions and answers from unstructured data. Led by founder and CEO Changsu Lee and based in San Francisco, the company has acquired over 100 enterprise customers in the US, Japan, and Korea, including SMBC, Persol, Kao, JPower, KB Securities, Hyundai Card, and Smilegate. In August 2021 it raised $10M in a round led by Atinum Investment with participation from Stonebridge Ventures, Sumitomo Mitsui Banking Corporation and existing investors, bringing total funds raised to $15M since its July 2017 founding. The company plans to use the proceeds to accelerate research and development, expand globally, and increase headcount in 2021. Allganize develops deep learning–based natural-language understanding (NLU) and conversational AI for enterprise customers. Its Alli.ai conversational AI and chatbot service aims to deliver highly personalized customer experiences by recognizing complex named entities and user intents from chat and voice data. The company leverages proprietary high-performance NLU technology to help companies build their own AI systems and services. Allganize has demonstrated market traction, securing business with Persol Process and Technology in Japan. The company intends to use the Series A proceeds to accelerate distribution of its solution to businesses. It operates from offices in Oakland, Seoul (Gangnam-daero, Gangnam-gu), and Tokyo.

Participated · Series A · Mar 2023

Indent provides VREVIEW, an AI-powered video review marketing tool that uses a chatbot to collect customer video reviews and product ratings which merchants can upload to their sites. The company says consumers using its video review platform showed, on average, a 6x higher order conversion rate in tests. Indent reports about 40 clients across the U.S., China and Japan and works with 3,800 online merchants, primarily direct-to-consumer brands; South Korean companies including LG Uplus, AmorePacific and MLB Korea also use its services. It claims more than 12 million registered users, roughly 60% of total online shoppers in South Korea, and says its data allows it to recommend about 66 million products. Indent is launching a B2C, TikTok-like service called Spray (in beta) that lets consumers upload monetizable video reviews and provides product recommendations to merchants and influencers. The company generates revenue through VREVIEW subscription fees and licensing fees when D2C brands buy user-generated content, and it employs 27 people in South Korea.

Participated · Series D · Oct 2022

한국신용데이터 provides payment rails and POS solutions targeted at small-business owners and has built capabilities considered essential for its merchant customers. Over the past two years the company steadily raised more than ₩2,000억 and completed a Series D totaling ₩2,400억. Management says it invested aggressively in payment infrastructure, POS solutions, and talent acquisition, expanding headcount by more than 2x. Those moves supported rapid commercial growth: the company reports growing revenue more than 20-fold. The firm emphasizes continued customer focus and cautious execution, noting plans to keep helping merchant customers through the coming second half. The company framed its fundraising and operational choices as deliberate investments made while other firms pulled back. Korea Credit Data offers a bookkeeping super app for small and mid-sized enterprises called Cash Note (launched 2017) that aggregates cash flow, credit card sales, expenditure, sales ledgers and policy information. The company has added SME-focused lending services that connect merchants to working capital loan programs. KCD claims more than 1.7 million registered merchants and has grown its app into a widely used platform among small business owners. The startup has made strategic acquisitions including ImU (POS) and Persona (government subsidy alerts), and it has announced the acquisition of Fiserv Korea as part of a partnership. KCD plans to use the new capital to expand its 230-person team and pursue additional acquisitions. Financially, the company has raised about $112 million since its 2016 founding and its Series D has reached roughly $70 million after the latest extension, valuing the business at about $776 million. Korea Credit Data Co., Ltd. announced a KRW 40,000,000,000 funding round that closed on April 20, 2021. The transaction included participation from Pavilion Capital Pte. Ltd., KB Financial Group, GS Holdings, Kakao Ventures, and KT Investment. Pavilion Capital invested $20,000,000 (KRW 22,345,400,000) as part of the round. The company has raised KRW 60,000,000,000 in total to date. Post-closing, the enterprise value of the company exceeded KRW 400,000,000,000.

Participated · Series A · Aug 2015

Jibo makes a consumer "social robot" designed to serve as a family-facing smart assistant. The device is intended to handle reminders, order food, take photos and video, send email, connect with internet-of-things devices, and facilitate communication. Jibo began as an Indiegogo project that raised $3.7M and offers an SDK for third-party developers to build features and applications. A full retail-ready version was not yet complete at the time of the article—the prototype could communicate but was functionally limited—and both developer and consumer units were sold out. The company ships to more than 45 countries including the U.S., EU and China and planned to start shipping developer units this fall, with earliest Indiegogo backers receiving orders that winter and later pre-orders arriving in late spring 2016. Jibo was founded by MIT graduate Cynthia Breazeal and does not currently plan an enterprise version. Jibo produces a family-focused social robot that uses artificial intelligence to engage users and serve as a home engagement hub. The company launched last year with a major media push and a successful Indiegogo campaign, taking in roughly 4,800 orders. It also raised about $2.3M via crowdfunding with roughly 5,000 backers. With the new capital Jibo plans to fulfill Indiegogo orders, hire staff aggressively, develop new prototypes, and grow a developer community to build apps for the platform. Initial applications include video chat, a security feature in development, a reminder/list-maker and a storytelling app. Leadership is shifting: founder Cynthia Breazeal moves to chief scientist and Steve Chambers becomes CEO. Jibo is a chipper, 11-inch-tall social robot prototype designed for families, featuring cameras, voice recognition to distinguish individuals, and touch sensitivity. It connects to a cloud platform to store family photos and is designed as an open platform that supports third-party "skills," though its own code is closed. Planned features include telepresence video calls with tracking, handsfree photo and video capture with smile/pose recognition, individualized message and calendar reading, and a storytelling function that combines graphics and physical movement. The developer edition is priced from $599 and aimed to ship in September 2015; the consumer unit is expected to cost around $500 with a consumer release slated for December 2015 and a full public release in early 2016. The project is led by MIT professor and social robotics specialist Dr. Cynthia Breazeal. At the time of reporting, the campaign had raised $356,000+ on Indiegogo and had exceeded its original funding goal by more than 3.5x, while TechCrunch had requested further details about Jibo's privacy policy and data use.