Hi, I’m Pranay — I build open-source AI that runs offline, learns fast, and puts devs back in control.
I believe the intersection of AI × Crypto is where the hardest, most meaningful problems live.
I build because I want devs and small teams to own their models, their data, and their infra — not rent them behind black-box APIs.
I care about making AI local, auditable, and fast — and crypto the backbone of trust and verifiability.
So I’m shipping:
- Yudai v3 — a programmable codex that turns product context into testable GitHub issues and PRs — crypto-native and local-first.
- solo-server — run open LLMs like Qwen and DeepSeek on your own machine, no middlemen.
- DeepSeek-R1 Distillation — proving reasoning doesn’t need massive clouds — just smart research + open hardware.
I’m here to help builders stay independent — and to push AI and crypto to serve people, not gatekeepers.
Practical AI for US/EU/Canada SMBs: deploy a private copilot, a do-the-task agent, or a tuned model — usually in weeks.
- AI Deployment (RAG & Small Models)
Private, brand-tuned assistants grounded in your docs, tickets, and data. Options: local/VPC or managed. - Agentic Workflows
Bounded agents that actually finish work (refunds within policy, scheduling, ticket triage, lead outreach) with audit logs + guardrails. - Fine-Tuning & Evaluation
Domain-tuned SLMs/adapters that beat generic LLMs on your tasks (lower latency, lower cost). Clear win-loss vs baseline. - Classical ML & Data Science
Forecasting, lead scoring, segmentation, anomaly detection — notebooks → services with monitoring.
Operations: document AI for invoices/POs/contracts, inventory signals, back-office runner.
Sales & Marketing: lead-gen copilot, brand-tuned content + ad draft + A/B loop, site FAQ/RAG that converts.
Customer Service: Tier-0 deflection bot, after-hours agent (status/refunds), voice inbox summarization.
Software & IT: repo-aware dev copilot, issue-to-PR pipeline, internal code/docs search with pgvector
.
- Discovery (free 30 min): pick 1–2 high-ROI workflows.
- Design Sprint (1 week): model/tool choices, evals, guardrails.
- Pilot (2–4 weeks): dockerized delivery, metrics, “go/no-go”.
- Scale: monitoring, retraining cadence, playbook handover.
Promise: If we don’t see a credible ROI signal by week 2, you pause — no hard feelings.
- Llama Impact Grant Winner — recognized for pushing open-source AI tooling (announcement)
- solo-server OSS Maintainer — powers 300+ indie dev deployments for local LLMs
- Yudai v3 — cloud-native + local codex chaining PM → Architect → Coder agents to ship test-first PRs
- Kernel KB8 Founder & Community Mentor — Gitcoin’s top 50 global founder cohort driving AI × Web3 innovation
- Web3 Infra Contributor — protocol tools for Mode, FortyTwo Money, EigenLayer, MegaETH testnet
- Finalist, MEGAZU Pop-up City — prototyping cutting-edge Web3 infra
- National-Level Hackathon Mentor — 50+ teams; winners at Smart India Hackathon & Prayatna 2.0 (AITR)
- Petabyte-Scale ETL @ CoinSwitch — Spark & Airflow for ML + risk pipelines
- Vgyaan (pre-GPT) — BERT-powered edtech that resolved 120k+ student questions/night
- I ship → learn → repeat 👷♂️ → 🚀
Core papers & concepts shaping Yudai v3 and my agentic stack:
- DeepSeek-R1 — RL + “bag-of-narratives” distillation for efficient reasoning
- Reinforcement Learning with Verifiable Rewards — provable reward guarantees for reasoning agents
- Absolute Zero Reasoner — pushing zero-shot/zero-reasoning capability in LMs
- SWE-smith & debug-gym — scaling bug synthesis + tool-augmented SWE agents
- LLM → SLM Agent Conversion — turning generalist LLMs into specialist SLMs for codex workflows
- DSPy — compositional orchestration for multi-agent pipelines
- NoFeeSwap Yellow Paper — zero-spread AMMs & liquidity design (I’m prototyping this in Solidity)
- Yudai v3 — invite-only pilot cohort rolling out now
- solo-server upgrades — smaller, faster, edge-ready models
- New Codex Agents — task-specialized SLMs, verifiable rewards, test-first PR workflows
Buyer Notes for SMBs (quick answers)
- Privacy & hosting: Local/VPC by default; managed is optional. Logs scrubbed; retention controls available.
- Integrations: Postgres/pgvector, Sheets, Notion, Slack/Teams, Gmail, GitHub, basic CRMs/helpdesks.
- Timeline: Most pilots 2–4 weeks. We start small, instrument, and scale what works.
- Pricing: Pilot (fixed scope) → Production (subscription + hours) → Platform (managed VPC).
- Fit: E-commerce/DTC, agencies/professional services, B2B SaaS, clinics/practices.