Bangalore · IST
Hi, I’m Archit Sharma. I build and run distributed systems.
Performance engineering at Red Hat from 2014, blockchain infrastructure since 2018, and now a second brain at Kurv Tech.
Building a second brain at Kurv Tech, which I started in 2025.
It is my full-time work. Until 2026 I freelanced as an infrastructure lead and fractional CTO, through Toptal and with other teams.
I started in performance and scale engineering at Red Hat (2014–2018), on the Cloud Suite: OpenShift, OpenStack, RHV and Satellite. I wrote Sarjitsu to make sense of SAR data.
Then came startups. I led engineering and consensus R&D for a public blockchain, led engineering and infrastructure for a computer-vision product that diagnoses diabetic retinopathy, and ran site reliability for another public blockchain. From 2022 to 2026 I freelanced through Toptal on DeFi, DePIN and L1 infrastructure.
In 2025 I started Kurv Tech, where I am building a second brain: software that remembers what you read and acts on it. It grew out of second-brain systems I built for my own use from 2021, alongside private prototypes and freelance work. It is now my full-time work.
- 2025–26 aPriori DevOps lead
- 2025–25 Monad DevOps
- 2025–now Kurv Tech Founder
- 2024–26 Velvet Capital Infrastructure lead
- 2023–23 Replicated Customer reliability engineer
- 2022–26 Toptal Infrastructure, SWE, team lead and fractional CTO (freelance)
- 2021–25 Independent R&D Second-brain systems for my own use, alongside prototypes and freelance work
- 2019–20 CasperLabs Site reliability engineer
- 2018–19 Rayd8 Chief software architect
- 2018–18 TrueChain R&D consensus lead, head of engineering
- 2014–18 Red Hat Associate performance engineer → performance engineer, R&D
- 2013–13 CERN openlab Summer student
- 2020-02-07 GRA FCUK? NAh
- 2017-04-30 Recovering encrypted LVM based partitions, after accidentally overwriting 'em with 'fio'!
- 2017-03-02 performance engineering TILs - part 1
- 2017-02-22 resizing an xfs partition on a thin provisoned disk in a RHVM env
- 2017-01-25 never go back, but never forget
201 issues and PRs across 58 projects, 68 merged.
- about · now · work the tabs above, by name
- writing · connect latest posts; email and elsewhere
- open-source contributions, with stats
- cv open the CV
- search <words> search everything (also ⌘K)
- theme flip light and dark
“In my quest for an answer to nature's existence, I wish to be an inter-galactic hitchhiker.”
Don’t panic.
Now
Kurv Tech
I started Kurv Tech in 2025 to build a second brain that people and their agents can share: a workspace for what you know, memory and a harness for your agents, and a crew that does the work. Every memory keeps its source and its owner, so you can check where an answer came from and who can see it.
- Kenby Early access The second brain you work in A second brain for people and teams: notes, a company wiki, investigations and R&D in one place. Every claim in an answer links to its source, and the sources are still attached when you come back weeks later. Works with Telegram and LinearEach source is yours alone or sharedInvite only for now kenby.ai ↗
- Temmo In development Memory and harness for agents A CLI that gives coding agents memory between sessions. Each session starts with what was decided before, and every memory keeps its source and its date. Runs inside the coding agent you useOld versions of a fact are keptInstaller not live yet temmo.run ↗
- Elvs In development The agent crew Small workers you reach from chat or a ticket. Each one works in its own closed-off box and hands back a draft. Nothing merges until you approve it. Skills are plain files you can readA receipt for every jobWorks with or without Kenby elvs.run ↗
Kurv Tech is my full-time work. Until 2026 I freelanced as an infrastructure lead and fractional CTO, through Toptal and with other teams.
Selected work
Built, deployed and led
- ~100×
- Ethereum’s throughput, with TrueChain’s hybrid consensus
- 5 → 100+
- validators on the CasperLabs network
- $25k/mo
- saved for io.net with self-hosted logging
- 2K → 5K
- nodes on a DePIN storage network
-
Trading infrastructure
A perpetual-futures exchange: C++ matching engine, order management and market data on multi-region AWS.
-
Blockchain networks
Validator and RPC fleets from testnet to launch, for Monad, aPriori, CasperLabs and TrueChain.
-
DePIN and the edge
A storage network grown from 2K to 5K nodes, and atomic OTA updates across 100+ edge machines.
-
Agents and memory
An agentic research pipeline: LangGraph, Neo4j and Qdrant retrieval, tools sandboxed in Firecracker.
-
Performance engineering
Scale tests for OpenShift, OpenStack and RHV, KVM I/O latency analysis, and tools like Sarjitsu.
-
Cost and reliability
Self-hosted observability that saved io.net $25k a month, and DDoS defences for Velvet Capital.
Career
My career as a trace
Every job on one time axis. Click a span to open it.
-
- service
- cern.openlab
- role
- Summer student
- period
- 2013 → 2013 (3mo)
Built enhancements to the CERN Document Server (CDS).
- Won Webfest ’13 (data visualisation) with Popular Physics, a citation-based physicist visualiser
- Part of ProtonMail’s early product brainstorm team
- Also interned at Red Hat (Pune, 2013: log analysis and an NLP search prototype) and TakeZero (2014)
-
- service
- redhat.perf
- role
- Associate performance engineer → performance engineer, R&D
- period
- 2014 → 2018 (4.3y)
- stack
- Python · Ansible · Elasticsearch · Grafana · pbench
Scale benchmarks across OpenShift, OpenStack and RHV.
- Led CloudForms performance, including the I/O impact analysis of the Meltdown/Spectre fixes
- Wrote Sarjitsu, a Linux monitoring and analytics framework, listed in the sysstat FAQ
- Added nested-type bucket aggregation for Elasticsearch to Grafana (merged), and wrote the collectd plugin for sosreport
- Benchmarked deep-learning libraries on RHEL across OpenBLAS, ATLAS and Intel MKL backends
- Co-authored the Satellite 6.2 performance and scale brief
- Presented at Red Hat Summit and DevConf Europe
-
- service
- truechain.consensus
- role
- R&D consensus lead, head of engineering
- period
- 2018 → 2018 (8mo)
- stack
- BFT · PoW
- link
- arxiv.org/abs/1805.01457
Co-designed and prototyped a hybrid BFT + PoW consensus at ~100× Ethereum’s TPS.
- Co-authored the TrueChain paper (arXiv, May 2018); TrueChain later patented the design (US 2020/0026699 A1)
- Led a 20-person team of distributed-systems engineers and researchers, within a ~50-person open community that included contributors from Google, Red Hat and VMware
- The design included a VRF-based rotating validator committee
- Led the testnet launch (Nov 2018, 1,200–1,900 TPS); mainnet followed in March 2019
- Secured compliance with China’s Ministry of IT; ran investor fireside chats, workshops and hackathons in Boston, Beijing and San Francisco
-
- service
- rayd8.platform
- role
- Chief software architect
- period
- 2018 → 2019 (7mo)
- stack
- Python · Flask · GKE · GPU
API-first product for diagnosing diabetic retinopathy from retinal scans.
- Generated K8s configs for GPU inference pipelines
- Trained computer-vision models; ran ML infra on GKE
-
- service
- casperlabs.sre
- role
- Site reliability engineer
- period
- 2019 → 2020 (1.4y)
- stack
- Terraform · Ansible · Rundeck · Python · Drone CI
Scaled the validator network from 5 to 100+ nodes.
- Second SRE hire; defined SLIs/SLOs for multi-region testnets; Casper mainnet launched in March 2021
- Ran the public web estate (site, blog, metrics dashboard, block explorer) and releases to GitHub, npm, crates.io and PyPI
- Built fault-tolerance testbeds for liveness, safety and throughput
- Ran CI/CD on Drone CI and GitHub Actions, with AWS and Nginx load balancing; revived the Drone publish-to-Bintray plugin for internal releases
-
- service
- independent.rnd
- role
- Second-brain systems for my own use, alongside prototypes and freelance work
- period
- 2021 → 2025 (4.0y)
- stack
- Python · Go · Kubernetes · Neo4j
Built and ran second-brain systems that I used every day, while prototyping MoaRAG and other private takes on a visual knowledge OS, and freelancing.
- AWS Activate (2021), Google for Startups (2022) and YC Startup School (2021)
- 3rd place at the KubeCon EU 2021 CNCF BugBash; the prize credits paid for CKA, CKS and CHFD (2022)
-
- service
- toptal.freelance
- role
- Infrastructure, SWE, team lead and fractional CTO (freelance)
- period
- 2022 → 2026 (4.1y)
- stack
- Rust · Go · Kubernetes · Terraform · Grafana · Loki
DePIN, DeFi and AI infrastructure through Toptal’s top-3% network.
- DePIN (fractional CTO): rescued an at-risk, seed-funded project; brought its launch forward by six months, grew the network from 2K to 5K nodes and cut node costs by 70%
- DePIN (fractional CTO): hired PMs, ran budgets, negotiated ~$350K in vendor contracts, and led a $20M contract ($2M in year one) with a European research consortium
- io.net (DePIN, decentralized GPUs): saved $25k/month by moving logging from SaaS to self-hosted ClickHouse and rook-ceph on EKS (50–100 TB, with failover)
- io.net: led a 5+ person DevOps team inside a 40+ engineer group through the airdrop rewards program and a Binance coin launch with $1B in participation; migrated Azure → DigitalOcean → AWS
- Also: Ray GPU inference on GKE (Lighthouse Technologies), a FastAPI app that launches EMR research clusters (DLT Science Foundation), Hyperledger Fabric for art provenance (Sdamp), a Strapi and Meilisearch migration (9thCO), Grafana dashboards (Rollee)
- Wrote the main whitepaper for a new data-availability layer
-
- service
- replicated.cre
- role
- Customer reliability engineer
- period
- 2023 → 2023 (5mo)
- stack
- Kubernetes · Go · Velero
- link
- github.com/replicatedhq/troubleshoot/pull/1366
Kubernetes software for customers in cloud, on-premises and air-gapped environments.
- Follow-the-sun on-call and support retrospectives
- Wrote the Velero analyzer for the open-source troubleshoot tool
-
- service
- velvet.infra
- role
- Infrastructure lead
- period
- 2024 → 2026 (2.5y)
- stack
- Go · gRPC · AWS · Cloudflare
DeFi infrastructure: security, scale and low-latency trading.
- Stopped DDoS attacks with WAFs and rate limits; re-architected across six AWS regions
- Go/gRPC intent solvers for low-latency DEX trades
- Active through 2025 and into 2026
-
- service
- kurv.tech
- role
- Founder
- period
- 2025 → now (1.8y) · running
- stack
- Go · Python · Rust · TypeScript · k3s
- link
- kurv.tech
Building Kenby, Temmo and Elvs: memory, agents and connectors.
- Kenby in invite-only beta at app.kenby.ai, built solo on a self-run k3s cluster
- Grew out of the MoaRAG prototypes (2021–2024)
-
- service
- monad.devops
- role
- DevOps
- period
- 2025 → 2025 (4mo)
- stack
- Kubernetes · ArgoCD · Grafana · bare metal
Testnet infrastructure for a high-performance L1.
- CI/CD and observability from scratch (ArgoCD, Helm, Loki, Grafana, VictoriaMetrics)
- 150+ node bare-metal testnet of validators and RPC nodes, with zero-trust access
-
- service
- apriori.devops
- role
- DevOps lead
- period
- 2025 → 2026 (1.2y)
- stack
- bare metal · Kubernetes · Grafana
Staking and MEV on Monad; now Capricorn.
- Bare-metal validator fleets through several testnets and simultaneous launches
- Cut infrastructure costs
- Arbitrum Open House London (June 2026): 2nd place, $20K, for Capricorn’s liquidity pools for tokenized stocks on Robinhood Chain; I built and ran the price pusher and fixed last-minute data issues
On the record
Papers, code, talks and wins
Papers
- TrueChain: Highly Performant Decentralized Public Ledger
- Performance & Scale Tuning of Satellite 6.2 and capsules
- QEMU KVM disk I/O: native or threads?
- User collections and query visualizations (CDS)
Open source
Talks
About
Thirteen years of making systems behave
Red Hat from 2014, then startups: consensus R&D, ML infrastructure and site reliability. Freelance infrastructure lead from 2022 to 2026, now full-time on Kurv Tech.
- Certified
- CKS · CKA · CHFD · RHCE
- Spoke at
- DevConf · Red Hat Summit · Blockchain Connect · PyCon India
- Won
- CERN Webfest ’13 · Docker Hack Day ×2
- Based in
- Bangalore, IST
Writing
Braindumps and musings
41 posts since 2011, from two old blogs.
- GRA FCUK? NAh heXoughts
- Recovering encrypted LVM based partitions, after accidentally overwriting 'em with 'fio'! heXoughts
- performance engineering TILs - part 1 heXoughts
- resizing an xfs partition on a thin provisoned disk in a RHVM env heXoughts
- never go back, but never forget digital sweat
- #noteToSelf - fix MariaDB-selinux related errors (fails to restart after update) heXoughts
Say hello
Email or LinkedIn both reach me.
archit [at] pm [dot] me