Amplify 2024 Lookback
It was the best of times, it was the worst of times, but most relevant for this post, it was a great year for Amplify. From dozens of new investments and follow-on rounds, to 3 new team members (plus 3 babies), a new publication, a podcast, and 120 issues of the Projects to Know newsletter…well, we’ve been busy. But not too busy to recap it all for you here. So without further ado, we give you The Amplify 2024 Lookback.
New companies announced
Amplify is an early-stage venture firm, so most of the investments we made this year are still in stealth. But an exciting crew launched to the world this year:
.png)
Allium
Blockchain data is messy and hard to work with, and with adoption across institutions and companies growing, good quality data for auditing, investing and analytics have never been more important. Allium tames the chaos by ingesting, sanitizing, and standardizing all data from 56+ blockchains and 1000+ schemas. Check out Visa’s industry leading Stablecoin dashboard, powered by Allium data.
Chai Discovery
Despite the majority of the bio community believing that AI is the future of biologics, most drug discovery today is still shockingly manual and low tech. Chai is building GenAI-first drug discovery, starting with foundation models for molecular struture prediction across proteins, small molecules, DNA, RNA, covalent modifications, and more.
DatologyAI
With the astronomical increase in data and compute volume for pre-training, most companies still overlook the most important input to state of the art GenAI models: good training data. Models are what they eat, and Datology is commercializing decades of groundbreaking research in data curation to help companies curate better, more efficient datasets. Initial results for LLMs show >80% savings on compute, >8 point increases in accuracy, and <2x improvements in cost per query.
Effectful
TypeScript – extremely promising, but not everyone's favorite language quite yet. Effectful is working on next generation Typescript: maximum type safety, more composable code, error handling from the future, request tracing, and everything else you need to build more resilient apps. Check out the open source Effect project.
Lamini
For enterprises, AI automation is only useful when it's accurate. Lamini's platform enables you to reduce hallucinations, while deploying securely anywhere and keeping inference latency and cost low. Check out their recent release: agents precision-trained to make function / API calls.
Warpstream
Despite quickly becoming the ubiquitous way to stream data, Kafka is still an absolute nightmare to run – it’s expensive and requires a dedicated team with expertise. WarpStream is an Kafka compatible data streaming platform, rebuilt from the ground up directly on top of object storage: no inter-AZ networking costs, no disks to manage, and infinitely scalable, all within your VPC. WarpStream joined Confluent in September.
Featured follow-ons
Our founders continued to shine this year and raise follow on rounds from other top tier investors; and not just in AI.
Realizations
Covariant
Entered into license agreement with Amazon.
Warpstream
Turbine
Read all about it
We published a lot more stuff this year.
We launched Amplified, a new publication for technical founders figuring out go-to-market for the first time. It’s practical, focused content on common issues like when to price your product, common mistakes with early pilots, frameworks for recruiting, and what gets to the front page of HackerNews. We have an amazing roster of guest authors from technical companies, plus our own Build Team. Everything is tailored specifically to technical founders, and is (of course) free.
Our very own Barr Yaron launched Barrchives, a video podcast about how leading AI-first companies like Vercel, Replit, and Runway actually build and support their AI products.
This year Sarah Catanzaro cruised past 120 issues of Projects to Know. It’s the best way to learn about new, interesting projects (open source, academic, or otherwise) in data, infra, and AI.
New Amplify team members
We added 3 new folks to the Amplify team this year (although we had known them for quite some time already).
.png)
Grace Ge is a Partner on the investment team. Grace joined Amplify from Menlo Ventures, where she led investments in companies like Pinecone, Orb, and Eppo (which is also an Amplify company). Next year, look out for her new video series – Git With It – where she interviews founders about topics like BYOC and getting your first few customers for a technical product.
Neiman Mathew is an Associate on the investment team. Neiman joined Amplify from Schmidt Ventures, where he worked on funding datasets and AI models in areas beyond language, video, and vision — like robotics, math, and biology.
Andy Aguiar is a Sourcer on the Build Team, helping technical founders recruit top notch technical talent at the early stage. Andy joined Amplify after years of experience recruiting for companies like Weights & Biases, Meta, and Segment (where he met teammates Natasha and Nate).
And speaking of new people…congratulations to Sarah, Lenny and Carlie on their growing families this year. Rest assured Sarah already has baby Camilla doing backpropagation by hand.
A 60+ event year
Amplify put on a hair over 60 (yes, 60) events this year – it’s the most we’ve ever done and we’re extremely proud of it, especially with a small team. Technical founders and leaders gathered at:
Thematic dinners like our Austin devtools dinner, quarterly Models and Bottles get togethers, 2 SF database and infra dinners, a Seattle systems and databases dinner, an AI research dinner in Paris, a Bio / Chem dinner and meetup, and our SF robotics dinner.
Conference events (and dinners, we like dinner) at Kubecon Paris + Park City, Data Council, MLSys, COLM, ICML, and NeurIPS (where Sarah is always looking for papers for Projects to Know).
Events for people (that’s right) like product dinners in SF and NYC, Head of AI dinners in SF and NYC, and Sales Leader dinners in…you guessed it, SF and NYC. Plus multiple Engineering Leaders dinners in SF and a Data Leaders dinner too.
We did events all over the U.S., from NY to Austin, Boston to Seattle, Philadelphia, NYC, and of course, our home in SF. But we also went international and hosted in Tel Aviv, Paris, Vienna, and Vancouver.
Product highlights from around the portfolio
Amplify companies released a lot this year, from instant mobile observability to unified APIs for building microservices, to groundbreaking GenAI models for video and text.
Bitdrift released Instant Insights, a way to get instant visibility into your entire mobile fleet, no config required. Instead of the usual with mobile instrumentation – a few weeks to deploy, collect, and then finally sift through and analyze data – bitdrift’s dynamic control plane lets you do this, well, instantly. Check it out here.
Chainguard passed over 1,000 container images secured on their catalog and remediated >50,000 CVEs, which translates to something like >100 work years saved for engineering teams at their customers. Check it out here.
dbt Labs announced a bunch of exciting new stuff at Coalesce 2024 like Iceberg table support, a visual editing experience, and dbt Copilot.
Diagrid released Catalyst, a unified set of APIs for everything you’d need to build microservices, from workflow orchestration to state management and event meshes. It’s all built on top of open source Dapr. Check it out here.
Eppo released support for Contextual Bandits, a class of algorithms that help enable 1:1 personalization in your product. You supply actions and their contexts to the Eppo SDK, tell them about the business metric you want to optimize, and Eppo does the rest. Check it out here.
Fermyon released Spin 3.0, their latest version of the open source developer tool for building, distributing, and running serverless WebAssembly (or Wasm) applications everywhere. It features component dependencies, selective deployments, and an OTel integration – check it out here.
Foxglove released Foxglove 2.0, focused on observability for robotics. A big part of it is the new Foxglove Agent, kind of like you’re used to with Datadog; you install it on your robot(s) and it collects and selectively uploads multimodal telemetry. Check it out here.
Hex announced their new Explore UI, a way for anyone to explore data fast and intuitively without having to write code. It makes it easy to quickly slice, dice, and visualize – whether it's a one-off question, or the beginning of a deep-dive analysis. Check it out here.
Datology turned their research to reality and implemented a groundbreaking data curation pipeline. It speeds up training by >40x and can save up to 98% on compute costs, all while improving model quality. You can read Ari’s paper about semantic deduplication (one of the techniques they used) here.
Runway introduced Act-One – a new way to generate expressive character performances inside their Gen-3 Alpha model using a single driving video and character image. Plus, Gen-3 Alpha Text to Video is now available to everyone.
Modal did a lot this year, from introducing WebSockets to lowering GPU prices to adding the ability to select regions.
System Initiative officially launched their open-source platform for DevOps automation in GA. It’s an intuitive, powerful, and collaborative replacement for Infrastructure as Code that reduces feedback loops by up to 90% through real-time, reactive digital twins. Check it out here.
Luma unveiled Dream Machine, a groundbreaking AI model that generates high-quality, realistic, and fantastical videos directly from text and images, built on a scalable multimodal transformer to deliver physically accurate, action-packed scenes. Check it out here.
TigerBeetle expanded their deterministic simulator to run on 1000 (yes) CPU cores.
Hightouch released AI Decisioning: it uses continuous experimentation and machine learning to find the most effective way to engage every customer. Check it out here.
Scribe introduced ScribeAgent, a new family of proprietary LLMs achieving state-of-the-art results on multiple web navigation benchmarks. In partnership with their co-authors at CMU, they open-sourced all the code, along with a version of the model trained on open-source datasets. Check it out here.
SiMa announced MLSoC™ Modalix, the industry’s first multi-modal edge AI product family. SiMa.ai MLSoC Modalix supports CNNs, Transformers, LLMs, LMMs and Gen AI at the edge, and delivers industry leading performance – more than 10X the performance per watt of alternatives. Check it out here.
Apologies if we missed your release in here – there was a lot to cover! See you again soon.
.svg)