Sam Silvester
R&D for startups · Ex-Improbable
Hello!
I'm Sam Silvester, an experienced software engineer with a deep tech mindset who specialises in helping early-stage tech startups. I also work on my own technical and research projects and enjoy collaborating with academic and industry researchers.
Previously, I was a principal engineer at Improbable, where I worked across high-performance simulations, compression, game and simulation engines, low-latency networking, and ML and AI research solutions.
I'm based in London and I'm always looking for new introductions and collaboration opportunities. Please get in touch!
About Me
I started Silvester Labs, where I help early-stage startups solve real technical and scientific challenges and explore my own research projects. This mission is borne out of my learned experience that the best innovations are not birthed by the companies able to raise the most funding but by those that learn to embrace constraint.
I worked at Improbable, where I became a principal engineer. I transformed Improbable's virtual worlds technology into a platform suitable for low latency action games, which I validated by implementing the highest scale fully-dense shooter game ever built.
I was also the lead of the networking team. My team was responsible for:
- Cutting-edge novel simulation compression layer that achieves huge bandwidth reduction in real-time, allowing for thousands of players in dense environments.
- Distributed low-latency networking stack, enabling 10,000x the ops per second of traditional games through horizontal scaling and level-of-detail protocols.
- Custom ultra high performance Unreal Engine fork, reassembling UE5's core logic to scale to 20k players.
- Novel spatial audio stack that allows 20k players to talk to each other in real-time.
I also worked as:
- An R&D team tech lead, where I rapidly iterated on new rendering and networking techniques to hit our record-breaking scale.
- A customer engineer, embedded in indie and AAA studios to solve their problems and integrate their engines with Improbable's platform.
- A platform architect, driving the long term product and technical direction of the company.
I studied Computer Science at the University of Cambridge, focussing on computational neuroscience and how deep learning helps us understand the human brain.
I learnt to code as a child, created educational programming videos as one of the earliest Roblox ambassadors and independently developed and self published a video game.
Consulting
I offer deep technical consulting services, specialising in early-stage startups with deep technical problems. My working engagements are very flexible and tailored to the engagement - from short term spikes through to long term production delivery or part time technical oversight.
You can find my technical areas of expertise above.
Rapid Prototyping
A different angle you want to explore, or an unexpected deadline, but don't have the capacity or expertise? I provide rapid delivery of prototypes with novel solutions to your problems.
R&D
I work independently or with your team to bring a deep-tech, research-oriented approach, and deliver cutting-edge technology directly into your product.
Productionising and scaling up
I turn deep technical concepts and prototypes into fully-fledged products.
Venture Partnership
For the right team, Silvester Labs can engage as a highly involved technical partner, offering high levels of investment while leaving you with more flexibility than a typical technical co-founder relationship.
Projects
Multi-Omics Latent Space Explorer
An interactive tool for exploring high-dimensional multi-omics data through learned latent representations, allowing researchers to navigate and inspect a space of potentially-interpretable embeddings.
Research Collaboration
I'm currently exploring collaboration opportunities with academic and industry researchers and students. No funding involved - if you think my background could be helpful to you, please get in touch. My current research interest is in the interpretability of multi-omics, but I'm open to anything interesting.
- "Fully-dense" meaning that all players could actually see (and shoot) all other players at any time. No phasing trickery or spatial sharding. ↩
- This wasn't even close to capacity. If you'd like help achieving this type of scale for a project that might be worthy, please contact me.↩
- I even had dinner with the CEO (by winning it in a raffle). ↩