Sid
Co-founder & CTO of Chimère
I am building agentic systems that are autonomous in every sense — not automation waiting on a human to approve each step, but software with an identity of its own, able to hold something, earn, spend, and answer for what it did.
How the security puzzles, the contact tracing, the payment rails and the agent identity layer turn out to be one problem.
Sidhantha Poddar is a serial technology entrepreneur building the financial rails for autonomous AI.
He started coding before he knew it could be a job, drawn first to security — taking systems apart to find where they give, then closing the gap.
At 21, during the 2020 lockdown, he founded DataNest Ventures and led 12 engineers to build privacy-preserving contact tracing that outperformed Europe's DP-3T standard by 3× — a winning entry among 2,000+ teams at the European Commission's EU vs Virus hackathon.
He then spent two and a half years at McKinsey's QuantumBlack, where his work on AI for heavy industry won the firm's global award for innovation and client impact.
Today he is CTO at Chimère, where he built the settlement rails, the compliance engine and the agent-native layer end to end. A US client pays and INR lands in an Indian freelancer's account in 20 seconds at 1% all-in.
Alongside it he is building Neuralabs, a project giving AI an identity of its own on-chain, so that intelligence can own, earn and spend on its own behalf. Five hackathon wins in ten months, and grants from Coinbase, ICP, Constellation and others.
His thesis is simple: money should move as fast as the thing that decided to spend it. First for the Indian freelancer waiting five days to be paid — then for the machines transacting in microseconds.
Companies and long-horizon projects, ordered by where my attention currently sits.
Co-founder & CTO · 2024 — present
Instant cross-border payments for Indian freelancers. A US client pays in USD and INR lands in 20 seconds, at 1% all-in with no hidden forex markup. The hard part was never moving the money — it's the compliance India requires on every inbound payment, so purpose codes and FIRA are generated automatically instead of filed by hand. The same rails expose an MCP endpoint, so the payer doesn't have to be a person.
Read the full story →
Founder · 2024 — present
Tokenising intelligence. An AI today is a feature inside someone else's product: it is nobody and owns nothing. Neuralabs gives it an identity held on a blockchain and represents everything around that identity as tokens — the workflows, the models, the data, the infrastructure, and access to each. An AI that earns can buy data access or compute, or spend on improving itself.
Read the full story →
Agent tooling, developer tools, contact tracing, and motion capture built from a gravity vector.
Agent tooling
Coding agents are capable and forgetful. An issue and knowledge tracker they operate natively and humans supervise: every issue is a folder the agent writes into, rendered as a browsable site, so state lives on disk and sessions compound instead of restarting.
Current flagship
Developer tools
A debugger that is a notebook. VS Code's debug console takes one command at a time and remembers nothing, so this attaches a Jupyter notebook to the live session: cells run against paused program state, and the whole investigation saves to share or re-open.
Published on the VS Code Marketplace
AI for social good
Contact tracing where the data is public and the identities are not. Every user held a decentralised verifiable ID with their public key as the only identifier, and each device computed its own infection probability at the edge — 3× the detection rate of DP-3T.
2nd of 2,000+ teams at EU vs Virus
Sensors & spatial computing
Motion capture without a camera. Gyroscope, accelerometer and magnetometer fused against the Earth's own gravity vector as the reference frame, so limb position came from each sensor's orientation rather than from anything watching it — in real time, on a microcontroller.
VIT Hackathon 2019 · under ₹600 in parts
Notes on machine learning, decentralized systems, algorithms, and building useful technology.
Hey there! Are you interested in accessing and interacting with a blockchain network but don’t want to go through the hassle of syncing the entire blockchain? Well, you’re in luck! Introduci…
Abstract We have several machine learning models available today which work in a very different way and produce a different result. Considering a given data set some models might work for on…
Genetic Algorithms are one of a class of approaches often called evolutionary computation methods used in adaptive aspects of computation that exceptionally involve machine learning and othe…
Live public repositories, pulled directly from GitHub with a resilient static fallback.
Open-source work in progress.
A knowledge + task system designed for AI consumers, with human-readable docs as a first-class output — Astro-based engine, folder-per-issue tracker, Claude Code plugin (agent-ks).
Gopeed extension: download an entire HTTP folder tree at once (recursive directory-listing crawler), structure preserved.
parallel distributed genetic algorithm
Open-source work in progress.
Fast multi-account switcher for Claude Code, in Rust. Named profiles, per-terminal activation, shared history.
Keep exploring
The Peepal Tree of Knowledge turns skills into a living map rather than a flat progress-bar list.