Mrinal Wadhwa

Great piece by Emilia David on the LangChain open source integration ecosystem, the power of LangGraph, and our newly released LangGraph Platform Helps put everything you see us working in perspective Link: https://lnkd.in/ghajS7zx

100

5 Comments

We released an improved speech-to-speech model and new API capabilities to help build production grade voice agents. Highlights: - Realtime API is now GA - gpt‑realtime: our most advanced speech-to-speech model is more natural, expressive, and better at following complex instructions (even mid-sentence language switches) and tool calling - New integrations: support for remote MCP servers, image input, plus phone calling via SIP - Two fresh voices: Cedar and Marin - 20% price reduction https://lnkd.in/gnzHiNjd

866

23 Comments

GPUs have gotten faster, and models have gotten larger. But storage is still stuck in the past. Modern AI workloads need instant access to huge datasets. Instead, teams waste time on slow retrieval, idle GPUs, and costly pipelines. Archil is changing this. It delivers object-storage scalability at block-storage speeds. The platform allows for seamless, high-performance access to massive datasets. By eliminating cold starts and accelerating throughput 30X, Archil enables teams to fully utilize their infrastructure and move faster on training, analytics, and AI deployment. Archil’s founder Hunter Leath brings rare technical depth and operating experience from building Amazon EFS and optimizing Netflix’s cloud performance. This is a team building the foundational layer that AI workloads desperately need. I’m thrilled to lead Archil’s seed at Felicis with Nancy Wang and support their mission to reinvent cloud storage for the AI era. https://lnkd.in/gWueXhzz

65

2 Comments

Ensuring the Embedder (YC S25) agent generates valid, functional code across thousands of different MCU families and board configurations is one of our biggest engineering challenges. Relying solely on physical hardware for CI/CD isn't scalable. To solve this, we pair Embedder with Renode, a development framework that lets you simulate physical hardware systems. In this demo, we break down our internal validation workflow: - Context Injection: Pre-loading datasheets, schematics, and code embeddings. - Autonomous Build: Prompting the agent to "one-shot" complex firmware. - Simulated Verification: Spinning up a Renode VM, attaching simulated I2C sensors (BMP180/LSM), and verifying the firmware logic via a virtual UART console. This workflow allows us to run massive parallel benchmarks, ensuring our model performance is consistent whether we're targeting an STM32, an nRF52, or a RISC-V core. If you're interested in verified firmware generation at enterprise scale, we're running pilots starting Q1 2026.

162

7 Comments

It's official 🤖 I'm excited to announce Alloy, and our $4.5m pre-seed round led by Blackbird with support from Airtree, Xtal, Skip Capital and an incredible group of angels, leaders and founders of Tesla, Waymo, Carbon Robotics, Halter, Reach Robotics, Relevance AI and more. One of the strongest pre-seed rounds ever done in Australia and a testament to the calibre of the team and magnitude of the opportunity. At Alloy, we are building the system for robotics companies to reduce time spent on processing data by up to 90%, enabling 10x faster iteration. Instead of leaving behind 99% of their data in cold storage, teams can continuously improve performance. This is one of the most pressing problems in robotics: robotics teams are drowning in terabytes of data they can't efficiently search or analyze, forcing every company to build expensive custom infrastructure from scratch. And the problem only gets worse as you scale. When your robot encounters an issue, has it occurred before? How frequently? Will you know if it happens again? You’re often left with more questions than answers, and hundreds of hours of log files to scrub through. Companies like Waymo, Amazon and Tesla have deployed hundreds of engineers at this data infrastructure challenge. We are bringing best-in-class to the world. Alloy allows robotics teams to automatically organise, search & analyse across all types of multimodal robot data in one platform using natural language, without waiting for labels. Instantly summarise missions and share. Unprecedented levels of observability and data accessibility. We're currently working with pioneering robotics companies like Hullbot, Breaker and more, across oceanic inspections, defence AI, and maritime autonomy. Special thanks to Tom Humphrey at Blackbird, Jackie Vullinghs at Airtree, Fiona Leng at Xtal and Olivia Grivas at Skip Capital for believing in our vision and being great partners on this journey so far. This is just the beginning as we build the layer that is unlocking the next wave of scaled robotics companies. Day 1 at Alloy. More in comments.

995

159 Comments

Today, 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. I've spent two decades building production systems and observability tools. In that time, I've learned that operating software is where most engineering time goes: 3AM pages, the cost optimizations to hit margin goals, the laborious compliance questionnaires. Production is the hardest part of software engineering, and every engineer I talk to has lived this story. AI is helping us write more code than ever, but that code still has to be deployed, monitored, and debugged in production. We're scaling code generation exponentially without scaling our ability to operate it. This is the gap Resolve AI closes. We're building AI for prod: agents that work across code, infrastructure, telemetry, and knowledge. Agents that bring together the expertise of multiple teams in every execution, absorbing context and learning continuously. Our goal is to help every engineer operate production fluently. Seeing the impact that customers like DoorDash, Zscaler, and Coinbase achieve with Resolve makes this work worth it. I'm incredibly proud of the team we've assembled to do it. Read my full announcement, linked in the comments

605

151 Comments

We're excited to announce that Argus Systems has raised $3M to build data and evaluation infrastructure for robotics and physical AI! 🎉 From our team’s previous experience building Waymo’s Perception data infrastructure, we know first-hand how critical it is to curate high-quality data and rigorously test model performance for real-world deployment. Argus Systems is now bringing that discipline to robotics. We’re proud to be backed by our two co-leads, Norwest and Maple VC, as well as Flex Capital, Depth Capital Ventures, The Graduate Syndicate, Parallel VC, and our amazing angel investors. We're just getting started—if you're building the future of robotics or interested in joining us on our mission, let’s connect! 🚀

577

73 Comments

The robotaxi race just got real. And Tesla is no longer on the sidelines. On June 22, Tesla will launch its first robotaxi fleet, just 10–20 cars, tightly geofenced within Austin. It’s a small rollout, but a very strategic one. Here’s why that matters: Unlike Waymo, which runs expensive, sensor-loaded Jaguars, Tesla’s every new Model Y is already FSD-ready. No retrofits. No added hardware. Just software updates. If this works, Tesla won’t just be a player they could scale robotaxis faster and cheaper than anyone else. Waymo may have the head start with 10 million+ rides and dominance in SF and Phoenix but Tesla has the edge in manufacturing at scale. In Elon’s words: 1,000 robotaxis could be live in months. A million by 2026. India may not be testing AVs just yet but we’re far from out of the loop. 🔹 Tata Motors is already supplying EVs to Waymo 🔹 Tata Elxsi is powering global AV software stacks 🔹 And Indian talent is quietly helping build the ecosystem worldwide But there’s a question no one’s asking loud enough: What happens after autonomy? - These vehicles are electric, sensor-loaded, running 24/7. -They'll age faster, wear out sooner, and retire earlier. So when that first million robotaxis hit the end of their operational life - Where do the batteries go? The sensors? The lidar? The rare earths? Who’s tracing them? Who’s recovering them? Who’s turning that material into opportunity? Here’s where India has an edge. Platforms like Retrace, built from our EPR and circularity ecosystem, are already digitizing the recovery of plastics and electronics at scale. That same tech - digital IDs, traceability, verified material recoveryc an power the afterlife of autonomous vehicles too.' That’s the value chain no one’s thinking about yet. Recovering, reusing, and monetizing their materials doesn’t just reduce waste. It reduces global dependency. It turns risk into resilience. Waste into value. The future isn’t just about building smarter mobility. It’s about building regenerative ecosystems around it. Tesla Tata Motors Waymo #evs #entrepreneurship #innovation #circulareconomy #recykal

85

2 Comments

Announcing Exa 2.1! We're now the state-of-the-art search API for both fast and agentic search. "Exa Fast" is the best API to power real-time chat and voice agents. "Exa Deep" is the best API for more complex searches over people, companies, coding, etc. To do it, we: - scaled our pre-training and test-time compute by an order of magnitude - made a research breakthrough - radically improved our infrastructure for crawling the web Building a search engine is one of the hardest technical challenges in software. It takes years of research and development to design retrieval algorithms over an index that’s many petabytes. On that path, we’ve needed to build semantic+lexical databases from scratch and massive-scale crawling infrastructure. We’ve now rediscovered and then gone beyond retrieval techniques that only Google and Bing used to know. We’re particularly excited that we’ve discovered novel neural methods for search that continue to improve with scale. In other words, Exa 2.2 will be coming soon :)

258

19 Comments

OCR-tuned VLMs are getting really good AND cheap, really quickly This is a cool release from LightOn Good and cheap models are starting to saturate OlmOCR-bench. There's a lot of greenfield opportunity in curating an even more complex, diversified benchmark (lots of complex PDFs out there!) and seeing how the latest models do on that https://lnkd.in/gF-ZNVwa

228

5 Comments

Just rewatched Paz's presentation at CMX — one of the more thoughtful breakdowns I’ve seen on scaling community. She shared how she used Intros AI 🤝 to automate different types of member introductions, making it easy to create meaningful connections without manual coordination. One line stuck with me: “We connected hundreds of people across 12 countries in minutes.” As community building evolves, so should our tools. Automating logistics isn’t about removing the human — it’s about making space for deeper, more intentional engagement.

37

4 Comments

💡 AI doesn’t just change how code is written – it changes expectations and outcomes. In the final section of Jellyfish’s AI Adoption Guide, we explore how expectations for engineering leaders are evolving as AI becomes embedded across the SDLC. Because the organizations pulling ahead are treating AI as a transformation, not a tooling upgrade. Inside, you’ll explore: - Why AI adoption is quickly becoming a leadership mandate, not an option - How to balance ambitious goals with realistic expectations - What cultural investments are required to sustain long-term gains - How data can be used to align executives and engineering teams The teams that win in the AI era won’t just have better tools – they’ll have leaders who know how to guide change. Download today: https://lnkd.in/dut_6WpQ

19