The Venture Codex

Participated · Series B · Nov 2020

Abacus.ai builds an end-to-end AI/ML platform that automates machine-learning model creation while supporting power users and custom models. The platform offers vertically integrated AutoML solutions for common enterprise use cases such as churn reduction, lead scoring and personalization. It recently expanded into computer vision, adding image detection, classification and segmentation use cases. Abacus reports roughly 10,000 users on the platform who have created over 30,000 models; several customers, including 1-800-Flowers, Flex and Recorded Books, use its models in production. Founded in 2019, the company currently has 45 employees and plans to double headcount by year-end, and runs outreach programs (including a high-school girls internship) to improve diversity. The company positions its verticalized approach as a differentiator against major cloud vendors and AutoML competitors like DataRobot. Abacus.AI (formerly RealityEngines.AI) offers a platform that simplifies building AI models by automatically training and optimizing them and now also provides modular services to host and monitor customer models in production. The company today launched Abacus.AI Deconstructed, splitting its service into modules for hosted model management, model explainability and de-biasing, and a real-time machine learning feature store. Abacus.AI emphasizes a full-stack, self-service approach intended to compete with offerings from major cloud vendors and reports traction with small- and medium-sized companies while building out enterprise sales. The company plans to use new funding to increase R&D headcount and expand its go-to-market team from two to ten people to better serve large enterprises. It also intends to add support for more languages and additional machine vision use cases in January. Abacus.AI was co-founded by former AWS and Google executive Bindu Reddy. Formerly RealityEngines.AI, Abacus.AI launched its autonomous AI service into general availability to help businesses implement modern deep‑learning systems without doing the heavy lifting of model training and pipeline engineering. The platform handles data pipelines, model training and production deployment; the company worked with 1,200 beta testers and counts customers such as 1‑800‑Flowers, Flex, DailyLook and Prodege. Recent product additions include unsupervised learning tools for anomaly detection and tools for debiasing datasets (with experiments improving facial recognition performance on underrepresented groups). Abacus.AI also launched a community model showcase and a pricing model that charges customers only when models are put into production. Roadmap priorities include building more connectors to third‑party systems (e.g., Salesforce and Segment) and expanding pre‑built solutions for language understanding and vision. The company has hired additional research scientists and engineers and expects headcount to grow from 22 to about 35 by year‑end; total funding stands at $18.25 million. RealityEngines is an AI and machine learning research startup based in San Francisco. Founded by CEO Bindu Reddy, COO Chetan Rai and CTO Arvind Sundararajan, the company is developing methods to train accurate models from sparse and noisy data, apply automated feature engineering and neural architecture search, and integrate explainability and fairness into ML systems. Its technical approach includes generative modeling and transfer learning to enable organizations to build models from incomplete datasets, plus automated pipelines for ensemble model construction. The company intends to use recently raised capital to build out a research and development team. Financially, RealityEngines secured $5.25M in funding to support these R&D efforts. The articles do not disclose revenue or user metrics.

Participated · Equity · Jan 2014

Fon operates a global, crowdsourced WiFi network and reports more than 14 million hotspots in over 200 countries. Founded in February 2006 by serial entrepreneur Martin Varsavsky, who serves as CEO, the company partners with major telecoms including BT, Deutsche Telekom, KPN, Proximus, MTC, MWEB, Oi, SoftBank and Telstra. Fon maintains offices in Madrid, London, Tokyo and New York. The company positions itself as a global WiFi leader and is pursuing a global growth strategy. Its recent financing activity includes a growth capital loan intended to support that expansion. Fon builds a global crowdsourced WiFi network by enabling people to share part of their private broadband via devices like the Fonera router. It is developing a "social" Fonera integrated with Facebook that will let friends and retail customers access WiFi via likes/check-ins, and a separate shop-focused Fonera for stores. Qualcomm has entered a strategic relationship to integrate Fon into Atheros chipsets and the Atheros SDK so third parties can include Fon access. Fon reported just over 12 million hotspots worldwide, grew 50% that year, and claims about 10% penetration in markets like Belgium and the UK, with a target of 35 million hotspots by 2016. The company pursues carrier partnerships (for example BT in the UK) and is talking to U.S. carriers to expand into the U.S. and target business/enterprise users. Fon has raised nearly $72 million since being founded in 2006. Fon sells consumer WiFi routers that enable homeowners to share bandwidth as publicly accessible hotspots. The company positions that network to serve growing demand for mobile WiFi from smartphone and tablet users. The new financing is intended to support expansion in the US and other markets. Fon was founded in 2006 by serial entrepreneur Martin Varsavsky. Prior to this round the company had raised over €40M in earlier financings. The product and expansion focus aim to capitalize on increasing mobile data demand. Fon deploys consumer WiFi routers (Foneras) that let users share bandwidth with other "Foneros" for free roaming access or charge guests and retain 50% of fees. The company reports about 170,000 active routers worldwide, with leading markets in the UK, Japan, France, Germany, and the U.S.; the UK alone has more than 70,000 BTFon members. Fon partners with telecom operators (notably British Telecom and Neuf) to enable roaming broadband offers and is preparing a launch in Russia with telco Sistema. Product plans include a Fonera 2.0 router (with a USB port for media sharing) due in June and an 802.11n router slated for December. Financially, this latest infusion is part of a multi-round raise that has taken the company’s total funding to more than $50 million. The company intends to use the new cash to fund the Russia launch and the new router rollouts. FON sells a €34.44 ($48.70) WiFi router and its core product is a community WiFi model where customers who share their home connection gain access to other members' hotspots. The company claims roughly 500,000 members but reports about 190,000 active hotspots, suggesting not all members host a hotspot. FON’s distribution strategy relies on carrier and location partnerships to expand coverage; it has signed deals with chains, ISPs and operators such as Neuf Cegetel and Time Warner. The company is pursuing wider UK reach through a deal with BT, which has more than 3 million broadband customers and will invite them to join FON’s service. The announcement indicates a revenue-sharing arrangement with BT, though the terms were not disclosed, and questions remain about how FON will monetize the free router distribution. FON faces competition from mobile broadband providers and emerging technologies like WiMax that could limit demand for shared WiFi.