I'm a Product Manager who builds with AI — not just talks about it.
At Infinite Locus, I manage product across multiple concurrent enterprise client engagements spanning e-commerce, ERP, health-tech, and marketplace domains. I own the full lifecycle: discovery, PRDs, user journey docs, stakeholder reviews, and cross-functional delivery.
Outside formal PM work, I build open-source developer tooling — MCP servers that connect AI agents to real-world APIs — and AI-powered workflows that make the PM job itself faster and more rigorous.
- 🚀 Enterprise PM — shipped ERP systems, web platforms, and client apps end-to-end at Infinite Locus
- 🤖 AI-native workflows — built a 3-agent PRD pipeline (PM → CPO → Designer) and a code-grounded PRD methodology that catches false assumptions before stakeholder sign-off
- 🔧 Open-source builder — published MCP servers for Zoho Mail, MS Outlook, and HackerOne API
- 📊 Data-driven — SQL, Python, Power BI, ETL pipelines, and ML models; validated by real analytics projects
- 🎓 IIT Delhi — B.Tech Mechanical Engineering, JEE Advanced AIR 927
Open To:
Open To:
- Product Manager / APM Roles (Full-Time)
- B2B SaaS & Enterprise Tech (India)
- Open Source Collaboration on AI ToolingLanguages
Frontend
Backend & Databases
Cloud, DevOps & Tooling
| Domain | Proficiency | Details |
|---|---|---|
| Agentic AI Systems | ⭐⭐⭐⭐⭐ | Claude Code skills, autonomous agent workflows, tool-use orchestration |
| MCP Server Development | ⭐⭐⭐⭐⭐ | Building Model Context Protocol servers integrating external APIs (Mail, ERP) |
| LLM Engineering | ⭐⭐⭐⭐ | Prompt engineering, RAG pipelines, structured tool-calling |
| Process Automation | ⭐⭐⭐⭐⭐ | Business workflow mapping, ERPNext/Frappe customization & automation |
| NLP | ⭐⭐⭐ | Text classification, embeddings, semantic search |
| Machine Learning | ⭐⭐⭐ | Supervised/unsupervised learning, model evaluation |
| Data Engineering | ⭐⭐⭐ | ETL pipelines, vector databases, integration pipelines |
🔹 geopolitics-digest — Autonomous Daily Video Digest
A zero-maintenance GitHub Actions pipeline that every morning detects new uploads across a fixed set of YouTube channels, has Gemini watch each video and write a sourced summary, clusters the day's reporting into topics, and emails every subscriber their own personalised brief.
| Category | Details |
|---|---|
| Stack | Python, GitHub Actions (cron), YouTube RSS, Gemini API, Gmail SMTP |
| Scale | 5 channels, ~15 videos/run, multi-recipient personalised delivery |
| Performance | Deterministic RSS diffing (no LLM guesswork on what's new), round-robin summary budget, dual-key quota rotation with a deferral net |
| Security | Credentials live only in Actions secrets — a config guard rejects secret-looking keys; recipient addresses are never cross-exposed |
| Impact | Replaces a daily manual "check every channel and watch" routine with a single zero-touch morning digest; state.json + runs.jsonl committed back as an audit ledger |
| Repository | Private repository |
Two-pass design: per-video structured summaries, then one text-only synthesis call that groups the videos by topic and orders them by how many channels covered each — with an automatic fallback to a per-channel layout if synthesis fails. Anti-hallucination throughout: RSS is the sole source of truth for new videos, every figure is quoted verbatim or omitted, and every claim carries its source link.
🔹 zoho-mail-mcp — Zoho Mail MCP Server
A Model Context Protocol server exposing Zoho Mail as a native tool surface for Claude Code, enabling AI agents to read, search, and manage email through the Zoho REST API.
| Category | Details |
|---|---|
| Stack | Python, MCP SDK, Zoho REST API |
| Scale | 12 production tools covering full mailbox operations |
| Performance | Direct REST calls, no polling overhead |
| Security | OAuth2 token-based authentication with Zoho |
| Impact | Enables agentic email workflows directly inside Claude Code |
| Repository | View Repository |
Built as a clean MCP integration layer, translating Zoho's REST API into structured tool calls consumable by any MCP-compatible AI client.
🔹 ms-mail-mcp — Microsoft Outlook MCP Server
A Model Context Protocol server bridging Claude Code with Microsoft Outlook via the Microsoft Graph API, giving agents native access to mail, folders, and messages.
| Category | Details |
|---|---|
| Stack | Python, MCP SDK, Microsoft Graph API |
| Scale | Full mailbox read/write tool coverage |
| Performance | Graph API batching for efficient multi-item operations |
| Security | OAuth2 with Microsoft identity platform |
| Impact | Extends Claude Code into enterprise email workflows |
| Repository | View Repository |
Designed to mirror the same tool contract as other mail MCP servers, making the underlying provider swappable without changing agent-facing behavior.
🔹 Claude-Skills-by-manish31-oss — Custom Claude Code Skills
A library of custom Claude Code skills built around PM-driven, autonomous development workflows — from product requirement capture to structured execution.
| Category | Details |
|---|---|
| Stack | Markdown-based Skill definitions, Claude Code SDK |
| Scale | Multiple composable skills covering PRD, planning, and execution flows |
| Performance | Zero-runtime overhead — pure prompt/workflow engineering |
| Security | No external data access beyond explicit tool grants |
| Impact | Standardizes how product ideas move from note to shipped feature |
| Repository | Private repository |
Encodes product-management best practices as reusable, gated workflows so agentic execution stays aligned with real product intent.
🔹 insights — Frappe Insights (Contribution/Customization)
A working fork of Frappe's open-source Business Intelligence tool, used to explore and extend BI/reporting workflows within the Frappe ecosystem.
| Category | Details |
|---|---|
| Stack | Vue, Frappe Framework, Python |
| Scale | Full BI stack — data sources, queries, dashboards |
| Performance | Inherits Frappe's query-caching and background job architecture |
| Security | Frappe's role-based permission model |
| Impact | Foundation for ERP-integrated reporting and analytics work |
| Repository | View Repository |
Used as a base for understanding and extending BI tooling inside the Frappe/ERPNext stack that underpins production ERP implementations.
June 2026 – Present
End-to-end product ownership across multiple concurrent enterprise client engagements — e-commerce, ERP, health-tech, and marketplace domains.
- Built a 3-agent AI documentation pipeline (PM → CPO → Designer) that generates PRDs and clickable prototypes from a single brief
- Developed a code-grounded PRD methodology — validates every product claim against real platform code before stakeholder sign-off; caught a foundational false premise that had survived multiple prior reviews
- Shipped ERP systems, web platforms, and client apps; owned requirements, sprint planning, design handoff, and delivery
Product Management PRD Writing ERPNext/Frappe Saleor Shopify Claude Code
June 2025 – July 2025 · Noida
Built LLM-based automation pipelines for funding research and anomaly detection across B2B technographic datasets. Automated data validation with Python and VBA.
Python VBA LLM Automation Data Engineering
| Recognition | Details |
|---|---|
| JEE Advanced AIR 927 (OBC-NCL) | Top 0.5% among 200,000+ candidates · 2022 |
| Harvard Asia Conference | Selected among delegates from 40+ countries · 2023 |
| Foundation for Excellence Scholar | Merit-based scholarship for leadership & academics · 2022 |
| Tata Data Visualization Program | Business scenario framing and data storytelling · 2026 |
| MasterCard Consulting Program | Promotion optimization and data-driven decision-making · 2026 |
| J.P. Morgan IB Job Simulation | Financial analysis and investment recommendations · 2025 |
✏️ Add your real certifications here. None are on record for this profile, so none are listed as earned.
✏️ Add your real handles below — remove any platform you don't use.
Current Focus:
Learning:
- Advanced RAG Architectures & Vector Search
- Distributed Systems at Scale
Building:
- MCP Servers for Claude Code (Mail, ERP integrations)
- Custom Claude Code Skills for PM-driven Development
- Autonomous LLM pipelines on GitHub Actions (daily video digest)
Exploring:
- Agentic AI Frameworks
- ERPNext/Frappe BI & Reporting (Insights)
Open To:
- Product Manager / APM Roles (Full-Time)
- B2B SaaS & Enterprise Tech (India)
- Open Source Collaboration on AI Tooling
