About
Interested in high performance distributed storage, streaming, and processing.
Activity
772 followers
-
Allston M. reposted thisAllston M. reposted this❗New Harmonic Hot 25❗ Resolve AI joins Lovable as the only two startups to top the Harmonic Hot 25 twice! This comes after a $40M April extension of their Series A, this time at a $1.5B valuation. Up 50% form the February round. Salient jumps to second and our agent Scout believes they’ve surpassed $30M in revenue and are likely to raise early next year. Aaru and Simile land in 3rd and 6th. Both are building simulation platforms to take on legacy market research. Aaru raised in December, but X-following indicates interest from Benchmark, Khosla and Felicis 👁️ 👁️ Thousands of investors use Harmonic every day. These are the startups they’re watching most closely as we approach Q3: 🥇 Resolve AI - Spiros Xanthos, Mayank Agarwal 🥈 Salient - Ari M., Mukund Tibrewala 🥉 Aaru - Cameron Fink, Ned Koh, John K. 4️⃣ DualEntry - Santiago Nestares, Benedict Dohmen 5️⃣ Traversal - Anish Agarwal, Raaz Dwivedi, Raj Agrawal, Ahmed Lone 6️⃣ Simile - Joon Sung Park, Michael Bernstein, Percy Liang, Lainie Yallen 7️⃣ Periodic Labs - Ekin Dogus Cubuk, Liam Fedus 8️⃣ Brain Co. - Elad Gil, Jared Kushner Et al. 9️⃣ Higgsfield AI - Alex Mashrabov, Yerzat Dulat, Mahi de Silva 🔟 7AI - Yonatan Striem Amit, Lior Div See the full list in the comments.
-
Allston M. reposted thisAllston M. reposted thisYesterday, we announced our Series A. We launched Resolve AI almost 2 years ago with a big mission of automating production operations. It has been amazing to go from whiteboarding ideas in a WeWork to serving customers like Coinbase, DoorDash, and Zscaler. The challenge of building out a new category that finally solves the pain engineers have been facing for decades has been the most rewarding of my career. I am incredibly proud of the team we have built, composed of people who are both extraordinarily talented and accomplished, and also kind and fun. It’s exciting to hit a big milestone like unicorn status, but I am more excited about what’s to come. The opportunity is huge and Resolve AI is the company to take on the challenge. Thank you to the amazing teams at Greylock Partners, Unusual Ventures, Artisanal Ventures, A* and Lightspeed. Saam Motamedi, Corinne Marie Riley, Raviraj Jain, Sebastian Duesterhoeft, Elijah Conniff, John Vrionis, and Andy Price, to name a few, have been incredible partners. We’re hiring amazing AI researchers and engineers to continue the mission, DM me if you are interested.
-
Allston M. reposted thisAllston M. reposted thisThere is so much hype around AI coding tools, and rightfully so, but writing code is only one part of software engineering. The hardest part is running production. AI for prod combined with AI coding will take us to the next era of software engineering.
-
Allston M. reposted thisAllston M. reposted thisToday we're announcing our $125M Series A led by Lightspeed, with participation from Greylock Partners, Unusual Ventures, Artisanal Ventures, and A*. This round values Resolve AI at $1B, bringing our total funding to more than $150M. Resolve AI is AI for production, helping engineering teams run software as fast as they build it. As systems grow more complex, the hardest part of production isn't writing code. It's understanding what's happening across observability, code, deploy pipelines, infrastructure, and the tribal knowledge scattered in people's heads. Resolve pulls that context together, reasons through the evidence, and helps teams act faster. Coinbase has measured 72% faster time to root cause. Zscaler reduced the number of engineers required per incident by 30%. To our customers: thank you. Extraordinary enterprises like Coinbase, DoorDash, Zscaler, Salesforce, MongoDB, and MSCI Inc., to just name a few, have trusted us to be part of how they run production. These results belong to you and your teams, and we're honored to be building alongside you. To our investors: thank you for believing in AI for prod before it was obvious. Sebastian Duesterhoeft and Raviraj Jain at Lightspeed, Saam Motamedi and Corinne Marie Riley at Greylock, John Vrionis at Unusual, and Andy Price at Artisanal have been true partners. To our 125+ team members at Resolve AI: thank you for betting on this mission. You all left great opportunities to bet on something bigger at Resolve AI, and you're the reason any of this is possible. Your hard work, dedication, and commitment is what make Resolve AI special. This funding is a moment in time, not the destination. What comes next is a future where AI doesn't just respond to fires. It helps auto-contain and resolve routine production issues before customers even notice, with clear oversight and guardrails. The agent era will create far more software than the last era. We're building the foundation to run it, and the work is just getting started. I've left my full announcement in the comments. If you want to see what AI for prod can do for your team, sign up and start using our product today or book a demo on our website.
-
Allston M. shared thisAI gave us the horsepower. AI for Production is the next unlock 🔐 Resolve AI just raised $125M at a $1B valuation, led by Lightspeed with Greylock Partners, Unusual Ventures, Artisanal Ventures, and A*. I joined a few months ago and we're moving fast. We're hiring. DMs open, or come say hi at Thunderhill this weekend 🏎️ See Spiros Xanthos blog in the comments for what we've been up to!
-
Allston M. shared thisI've joined Resolve AI 💚 Building and operating production systems at scale is hard. New features that need to handle every edge case. Reliability despite unreliable dependencies. Debugging parts of the system that when you “git blame” were last changed 7 years ago is tough. It requires tribal knowledge & deep expertise across so many layers to solve problems efficiently. For years, I've wished for something that could help bridge that gap. Not just for SREs, but for anyone building production systems. Then, this all accelerates with the rise of vibe coding. That code is harder to review, but still eventually shipped and lands in production. The complexity compounds, your understanding degrades, yet you're still on-call for those 3AM pagers. Resolve AI is being built at the perfect time to solve this. It actually understands your production stack. The way it approaches observability and system intelligence feels like a genuine shift in how production can look. The team has lived and breathed the observability and agentic space, with the passion to make this really work. Already loving building alongside them 💪 Spiros Xanthos, Mayank Agarwal, Rushin Shah
-
Allston M. posted thisWrapped up 6.5 years at LinkedIn on Monday, my first real adult job. Thanks to everyone who helped me on that journey. Here are some takeaways for those starting their careers or anyone who finds them useful. - Communication matters more than engineers like to believe. Many of us love coding, but communicating your ideas and impact to various audiences is crucial. Some say communication matters most, others say technical accomplishments do. They're synergistic. - Find a company you like working for, both for the work and the people. It makes difficult times easier and fun times more fun. If you align with their culture, values, and morals, even better. I really enjoyed LinkedIn's Work-hard Play-hard culture, and providing economic opportunity for every member of the global workforce. It took me some time to embrace the other values like Relationships Matter (and they really do). - Every technical decision has history and context. It's easy to criticize something without realizing it addressed a specific problem under constraints like limited headcount, a tired oncall emergency at 3AM, or accepting that a better solution just wasn't worth the investment. There aren't many right or wrong decisions, just better or worse ones at that point in time. A rearchitecture 5 years later, or fixing a bandaid created 1 sprint ago, neither invalidate what was right then. - Build friendships and community. For me, this was shared interests in personal wellbeing (something LinkedIn really emphasizes) through gym sessions and obstacle races, plus organizing Friday fun drives and running into people at track days 🏎️ - You can leave an employer even if you like it there. Maybe you want something different, or the role was right then but isn't now. You and your situation change. Make staying a conscious choice, not a passive one. I checked in with myself every 6-12 months about whether I wanted to stay at LinkedIn, and for a long time it was a hell yes. - Small things compound daily. "Technical surplus" in your codebase (small improvements creating sustainability like investment returns, opposite of technical debt) or daily habits shape you as a person. My driving, friends, and community have unintentionally made me more risk-seeking, which seeps into career choices too. - Keep expenses low and cash flow high for as long as you can (no big loans, even if you can afford them). When expenses are low, you won't feel chained to your big tech job. Not everyone has this luxury, so do it while you can. Your expenses naturally grow when you're older, making it harder to save. Those responsibilities will cost you some individual freedoms, and you want to be intentional about that. I'm super grateful for my coworkers and friends who supported me throughout the years. I'm excited for what comes next 🚀
-
Allston M. posted thisCursor is a power tool for engineers that know what they want and how to build it, but simply don't have time. My workflow in the last week has changed to opening as many codebases that I have ideas on but no time to implement, running in agent mode in all of them, and basically doing code review the whole time. And this is so much more interrupt-friendly, like when helping teammates or debugging an oncall issue, the agents are all making progress. These tools are great for Senior or Staff engineers that traditionally have to scale up their team, which takes time away from their own coding. Now I have that coding time back, and it takes even less time than before. With 3.7 it was marginal, but with 4 it feels like the reality of having an orchestrator agent manage all my ideas is much closer than I thought. Cursor & Claude 4 is truly a step function improvement. With 3.7, I was on the second day of constant prompting to make incremental progress. 3.7 was getting stuck, required intervention (manual coding & debugging, the horror!), and even made itself lose its progress at one point, and it forgot how it moved forward. I restored a last-known-good checkpoint where 3.7 made a chunk of progress, and switched to Claude 4 once it was released. Claude 4, for the next 30 minutes, was fully autonomous from the initial prompt, and solved the problem 99% of the way there. Where on 3.7, I spent roughly 12 hours of continuous prompting, 4.0 basically solved it in less than an hour. This isn't just a 10% improvement from the benchmarks.
-
Allston M. posted thisI've been skeptical of all the AI tools, but today was the first time I used a proper AI agent in a development workflow, and it is an interesting perspective shift. I always asked Claude anything from coding, to systems design, to working on my car, to adjusting gym routines to work around nagging pains. The typical chat interface is excellent at what it is, but obviously can lead people down the wrong path, as its trying to aggregate scattered information. People must apply a lot of judgment to things like that which are harder to reverse, and hands-on or offline skills don't have as much online content as software. Most software is easy to revert. If an AI model told you something bad, or did something bad, you're usually fortunate enough to be able to revert it quite quickly. Not all industries can do that. I've had Agents: create 2 separate implementations in code and only use one of them (dead code), try calling code that didn't exist, delete code that was still used in the same repo, breaking changes in schemas without telling me. These are usually obvious to a human, but difficult for an Agent, even when it has access to a virtual environment. Yet, I saved at least 1-2 days worth of engineering effort by using the Agent. What would've taken me 3 days for a very practical industry issue, took me less than 1 day. Was the Agent wrong quite often? Yes. Did I need to start over from scratch? Yes. Did the Agent eventually solve the problem? Also yes. What's different, is the time it takes to write every single keystroke, read every single API documentation, understand how every piece glues together, and iterating in the normal dev cycle. I think there's a new dev cycle emerging with Agents. I'm not calling it "Vibe Coding". There was a lot of real eng effort and knowledge needed for it to be feature complete. But getting the Agent to do what you want, with significantly less effort than a human, is ultimately the goal. This development cycle is also iterative, but rather than you writing the code and testing, you're iterating on finding out what the Agent knows and doesn't know, and how to steer it in the direction you need it to. In some sense, it was very natural feeling because you can treat it like any other teammate: they have some knowledge in area X but not Y, and you need to teach (prompt) them smartly about Y. And the order of your prompts matters significantly. If you prompt too specifically, it might not take a broader look and fix issues elsewhere. If you prompt too generically, it might start playing a game of whack-a-mole with itself. There is a very real skill that we all will need to learn in how to communicate best with these Agents and models, and there's a lot of carryover to your communication with the people you actually work with. As an aside, it was awesome to just type simple things like "verify & push", and the Agent knows exactly what you wanted, with all the special arguments or flags.
-
Allston M. liked thisAllston M. liked thisQ3 is starting strong for Resolve AI with another Fortune 500 Enterprise becoming a customer (in fact a massive Fortune 100 company this time), who also told us that we are a year ahead of competition. We have been working with some of the most sophisticated technology companies, such as Coinbase, Snowflake, Doordash, Robinhood and others. They all have very deep adoption of AI across development and production and have helped us build the best AI SRE platform. But AI for Production and SRE will have even bigger impact to larger enterprise customers that are dealing with multiple generations of applications and infra, complex hybrid environments and manual operations. AI in general, and Resolve AI, is getting to the point that we can truly and measurably improve reliability and MTTR and also deflect/automate huge amount of work happening across L1-L3 ops. It took longer than coding because of the complexity of production systems but also the maturity it took in accepting AI in production but we are now starting to see massive ROI. A couple direct quotes from major enterprise customers: "The on-call model has flipped: AI-led, human-assisted (from human led, AI assisted for a few quarter before)" "As adoption of Resolve grew, major incidents went down. A sign that teams were catching problems earlier, before they escalated"
-
Allston M. liked thisAllston M. liked thisEvals. Evals. Evals. Hillclimb. Hillclimb. Hillclimb. -Dhruv Mahajan, Chief AI Scientist at Resolve AI Don't fall for the 80% trap. With a production agent, the hard part is the last 20%. You need a loop to keep those models improving, and that loop requires evals built against your own production. Creating the right evals is a core ML problem: easy to get wrong, easy to make too easy, and hard to know what you should be climbing toward. It's not a project you finish, and most teams aren't staffed or ready for that. Listen to the full piece here: https://lnkd.in/eSrGSEhZ
-
Allston M. liked thisAllston M. liked thisI'm excited to share that I've joined Resolve AI! Resolve AI is building agentic AI to help engineering teams run production more confidently in an increasingly complex world. The team, the technical depth, and the customer traction made this an incredibly easy yes. Huge thanks to Spiros Xanthos, Mayank Agarwal, Brooke Daniels, Vasi Karkantzos, Juan Campos and Sophie Tung for the warm welcome! Looking forward to learning, building, and helping define the AI for prod category. And to everyone at Headway who shaped how I think about AI over the last two years (you know who you are), thank you. I can confidently say it was one of the most formative stretches of my career and I will miss the team dearly.
-
Allston M. liked thisAllston M. liked thisExcited to be joining Resolve AI! What drew me in was energy in the room, the size of the problem we're solving, and the caliber of people I'd get to learn from every day. Resolve AI solves a real problem: helping engineering teams spend less time firefighting and more time building, so companies can move faster with confidence. Thanks to the team Spiros Xanthos Mayank Agarwal Chris Dyer for the opportunity and the warm welcome!
-
Allston M. liked thisAllston M. liked thisHow I sleep when I'm on-call knowing Resolve AI's got my back
-
Allston M. liked thisAllston M. liked thisI'm excited to be joining Resolve AI! We're building AI for production. I believe that production systems are becoming increasingly complex. The introduction of AI coding tools have largely eliminated the coding bottleneck (or we'll get there soon). But that bottleneck hasn't disappeared; it's just shifted to the right. The new challenge is handling the explosion of code in production, where systems are becoming increasingly complex to manage and maintain. We're tackling this problem head on at Resolve. Founded by the co-creators of OpenTelemetry and AI researchers behind Google Deep Research, Resolve AI is already a leader in using AI to manage production systems. Our customers run some of the largest and most complex distributed systems in the world and they're trusting Resolve AI to help them. Huge thanks to Spiros Xanthos Mayank Agarwal Daniel Pham Chris Dyer and Sophie Tung for the opportunity to build this alongside them!
-
Allston M. liked thisAllston M. liked thisNew chapter: I've joined Resolve AI as an Enterprise Account Executive! Resolve AI is AI for production — built to give engineering teams their time back. Today, so much of engineering capacity gets tied up in running production instead of building: chasing down alerts, piecing together context across a dozen tools, and manually investigating issues that could be understood in minutes instead of hours. Resolve changes that. It works alongside engineering teams, understanding your systems the way your best engineer does, so teams spend less time firefighting and more time shipping. Already trusted by teams at Snowflake, Robinhood, Zscaler, and Coinbase, and backed by top investors including Lightspeed and Greylock, with $190M raised to date. Thankful to Mayank Agarwal and Spiros Xanthos for the warm welcome. Excited to get to work!
-
Allston M. liked thisAllston M. liked thisI joined Resolve AI! Resolve is building AI that runs production autonomously so engineers can get back to building and innovating. As an investment banker at Goldman Sachs, I saw the caliber of focused strategy, relentless execution, and exceptional talent within the best-in-class generational companies. Resolve has all three. At McMaster-Carr, I led customer-facing teams at one of the most operationally rigorous companies. This gave me a ground-level appreciation for how deeply production reliability is woven into operations and the customer experience. Having seen that complexity, speed, and precision up close, I understand what's at stake. Tackling such a large opportunity is not only an interesting research problem but also a compelling commercial deployment problem. Resolve is already making an impact for customers across a wide range of industries, including software, healthcare, financial services, and e-commerce, with more to come! Many thanks to Spiros Xanthos, Brooke Daniels, Vasi Karkantzos, Josh Grose and Juan Campos for the opportunity. Excited to be part of building and defining this category!
-
Allston M. liked thisAllston M. liked thisIf you have the opportunity to work at an AI company and aren’t doing it, you’re missing the chance to experience first hand the biggest technological shift of our generation. It's hard and extremely demanding. I never worked this hard and I was a much less experienced founder in my last 3 startups - 2 of which got acquired by VMware and Splunk. But it will advance your career at 5x the pace. This is true for engineers and for GTM professionals alike. The way AI companies are selling, scaling and partnering with their customers has completely changed. I have learned more about sales in the last year than the last 10 years combined. Resolve AI is hiring in many roles in GTM: sales, technical sales, growth, customer success. I expect explosive growth in the 2nd half of 2026 and more so in 2027. The AI SRE/AI for Prod market is just starting to inflect as if took a lot more advancement in models than other vertical AI markets. Today we announced 6 hires in GTM and I couldn't be prouder of the team we are assembling. Aaron Daniel, Javiera Garrao, Nick Marshall, Stuart Maynard, Dylan Wood and Lukun Zhang welcome to Resolve AI!
Experience & Education
-
Resolve AI
View Allston’s full experience
See their title, tenure and more.
or
Licenses & Certifications
Volunteer Experience
-
Organizer & Software Development Lead
Santa Cruz Hack Inc.
- 10 months
Science and Technology
Developing various services for UCSC's hackathon for the 2019 school year. Involves application submissions and decisions, live schedules, SMS notifications, and other logistics. Primarily working on handling registration, acceptances, and check-ins of 1000+ students.
backend registration API: Flask, Gunicorn, PostgreSQL, Nginx, Docker, Heroku
frontend (https://cruzhacks.com): React, SASS, Heroku -
Mentor
HackSC Junior
- Present 8 years 5 months
Science and Technology
Helped boys and girls learn programming fundamentals at the University of Southern California.
Python, Scratch
Courses
-
Abstract Data Types
CMPS 101
-
Algorithm Design and Analysis
CMPS 102
-
Computer Architecture
CMPE 110
-
Database Systems I (Externals)
CMPS 180
-
Database Systems II (Internals)
CMPS 181
-
Distributed Systems
CMPS 128
-
Functional Programming (Haskell, Lambda Calculus)
CMPS 112
-
Intro to Computer Networks
CMPE 80N
-
Intro to Symbolic Logic
PHIL 9
-
Operating Systems
CMPS 111
-
Security
CMPS 122
-
Software Engineering
CMPS 115
-
Theory of Computation
CMPS 130
Projects
-
Budget It @ Stanford's TreeHacks 2018
See project- Web app with Amazon Alexa support to manage finances through smart budgeting by expenses.
- Implemented Amazon Alexa skill with Google’s beta database. Helped integrate with front-end.
- Received runner-up for Best Voice Experience using Amazon Alexa.
Python, Flask, AWS Elastic Beanstalk, GCP App Engine, GCP Firestore Beta -
Veost @ UCSC's CruzHacks 2018
- Face-to-face video chat web app to reduce stress, raise mental awareness, and receive help from others.
- Deployed web server and implemented back-end logic for user authentication.
Python, Flask, and JavaScript, HTML, CSSOther creators -
-
Systhetic
-
Aesthetically pleasing near real-time system monitoring. Like Grafana, but better.
- Architected system components such as monitoring, near real-time streaming, database, authentication, and displaying metrics
- Implemented near real-time data flow from server to browser
Demo: https://www.youtube.com/watch?v=g7oqqPvi5M4&fbclid=IwAR0fYXtTkHCOev3rRWKshkD8tmdVCwSZX6VH6GkaoMK88WEAu8LaI2Ypk0w
Go, React, Express, ChartJS, MySQL, AWS Kinesis Data Streams, Firebase…Aesthetically pleasing near real-time system monitoring. Like Grafana, but better.
- Architected system components such as monitoring, near real-time streaming, database, authentication, and displaying metrics
- Implemented near real-time data flow from server to browser
Demo: https://www.youtube.com/watch?v=g7oqqPvi5M4&fbclid=IwAR0fYXtTkHCOev3rRWKshkD8tmdVCwSZX6VH6GkaoMK88WEAu8LaI2Ypk0w
Go, React, Express, ChartJS, MySQL, AWS Kinesis Data Streams, Firebase Authentication, DockerOther creators -
-
Speculator
-
See project- RESTful API to predict price trends of cryptocurrencies with machine learning.
- Implemented technical analysis momentum indicators on historical price data from an exchange.
- Utilized random forests and deep neural networks to classify price movements and tendencies.
- Streamlined developer interfaces by refactoring machine learning models to identical APIs.
Python, scikit-learn, TensorFlow, Pandas, NumPy, Flask
Organizations
-
Association for Computing Machinery, UCSC Local Chapter
-
- Present
Other similar profiles
Explore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content