# Priyam Jain — Published projects
HTML project catalog: https://priyamjain.me/catalog/
Selected case studies and career profile: https://priyamjain.me/llms.txt
## AI Collaboration & Task Summarization Agents
Portfolio page: https://priyamjain.me/projects/?id=HSZU5rUAvPHLlUSR1ZK1
Category: AI Collaboration & Task Summarization Agents
Summary: A set of AI agents designed to summarize tasks, coordinate contributors, and maintain operational context across collaborative projects. These agents act as a glue layer between human work and structured systems.
Problem: In collaborative projects:
Context is lost quickly.
Tasks become ambiguous.
Onboarding is slow.
Coordination requires constant human overhead.
Solution: I built agents that:
Summarize tasks and discussions.
Maintain shared context.
Assist contributors proactively.
Normalize human workflows into machine-readable state.
Integrate with platforms like Pinchit and Opbase.
They reduce coordination cost while improving execution clarity.
Technology: LangChain · LLM APIs · Vector DB · Queue Workers · API Services
Tags: AI Agents · Task Management · Summarization · Collaboration · Automation
## AI-Based Kubernetes Anomaly Detection System
Portfolio page: https://priyamjain.me/projects/?id=zGVucLVXXLaMrSCOn6lA
Category: AI Infrastructure · Observability · DevOps · Reliability Engineering
Summary: An AI-powered anomaly detection system for Kubernetes clusters that identifies early failure signals, traffic spikes, and abnormal behavior in production environments to improve reliability and response time.
Problem: Production clusters generate massive telemetry, but:
Failures are detected late.
Engineers react instead of predict.
Alerts are noisy and shallow.
Solution: I built a detection layer that:
Learns baseline behavior of services.
Detects spikes and anomalies early.
Correlates metrics with events.
Surfaces root-cause context via AI summaries.
Enables proactive reliability engineering.
It shifts ops from reactive to predictive.
Technology: Python · Kubernetes APIs · ML Models · Metrics Pipelines · LangChain · Observability Stack
Tags: Kubernetes · Anomaly Detection · Observability · AI Ops · Infra
## Blockchain Escrow — Trustless Payment & Commitment Layer
Portfolio page: https://priyamjain.me/projects/?id=IaRdQp3kxFTmTRw2GmMj
Category: Blockchain · FinTech Infrastructure · Smart Contracts · Trust Systems
Summary: A blockchain-based escrow system that enables trustless collaboration between contributors and organizations. Funds are locked in smart contracts and released automatically based on proof-of-work, milestone verification, and dispute resolution logic, reducing counterparty risk in open collaboration platforms.
Problem: In open collaboration:
Contributors fear non-payment.
Clients fear non-delivery.
Manual escrow is slow and centralized.
Disputes are expensive.
Trust becomes a bottleneck for scale.
Solution: I built a trustless escrow layer that:
Locks funds via smart contracts.
Releases payments on verified milestones.
Integrates proof-of-work signals.
Supports dispute resolution flows.
Connects on-chain commitments with off-chain work systems.
It enables scalable collaboration without centralized trust.
Technology: Solidity · Smart Contracts · Ethereum / L2 · Web3.js · Backend Oracles · Event Indexers
Tags: Blockchain · Escrow · Smart Contracts · Web3 · Payments · Trust Infrastructure
## Cloudshield
Portfolio page: https://priyamjain.me/work/cloudshield/
Status: Production observability
Summary: Helped bring an AI-assisted troubleshooting platform for distributed infrastructure into production.
Role: Contributed to backend engineering, observability and production delivery for AI-assisted investigations.
Engineering impact: Helped bring diagnostic evidence and AI-assisted investigation into a workflow engineers could use in production.
Business impact: Gave engineers a repeatable way to assemble diagnostic evidence for production incidents.
Scope: Published at a high level to protect product and operational details.
## Agentic Lineage
Portfolio page: https://priyamjain.me/work/agentic-lineage/
Status: Open source
Summary: Created Agentic Lineage and lead its development. The open-source Go runtime packages and distributes agent workflows across coding tools.
Role: I conceived Lineage and lead its technical direction, roadmap and releases. I designed the package and execution model and documented a review standard that asks contributors to link an issue and explain how their change is tested. I also presented the project and discussed practical reuse with technical leads.
Architecture: A manifest declares package exports and capabilities. Deterministic archives carry the assets, and the receiver verifies content digests before use. The registry stores versioned artifacts while local provider adapters materialize the workflow for tools such as Claude Code and Codex.
Engineering impact: Made agent environments inspectable, versionable and reusable across projects, with provider-specific setup behind an adapter boundary.
Business impact: Technical leads discussed how to reuse shared workflows. The repository and releases show the delivered implementation.
Workflow: Package and publish a workflow, resolve a version, inspect its contents, then enable it and run it through a provider adapter.
Scope: The product exposes declared capabilities for inspection. Those declarations do not enforce runtime isolation.
Source code: https://github.com/agentic-lineage/lineage
Architecture: https://github.com/agentic-lineage/lineage/blob/main/docs/architecture.md
Pull request review standard: https://github.com/agentic-lineage/lineage/blob/main/docs/decisions/0004-require-issue-links-and-test-plans-for-prs.md
Releases: https://github.com/agentic-lineage/lineage/releases
## Opbase
Portfolio page: https://priyamjain.me/work/opbase/
Status: Operational use
Summary: Built an industrial order operations platform with more than six active clients, as reported by its operator.
Role: Designed the workflow and built the operational software linking order intake, catalog, inventory and fulfillment.
Architecture: NestJS and PostgreSQL support a shared order record and explicit transitions. React gives owners, store managers and order managers different views of the same work. Webhooks and notifications connect channel events and operational handoffs.
Engineering impact: Made the order record the place to check status, ownership and partial deliveries. That gave staff a shared record for day-to-day operations, documentation and analysis.
Business impact: Opbase has more than six active clients, as reported by its operator. In one operation using it, the team reported zero missed orders and zero missed compliance steps, with routine coordination down to a minimum. Annual documentation and analytics work fell from several weeks to about two days.
Workflow: Capture an incoming order, assign it, track status and partial deliveries, and surface open items to the next owner.
Scope: The active client count is a firsthand report from November 2026. The operating improvements describe one deployment.
Opbase overview on Pinchit: https://www.pinchit.in/#work
## Pinchit — Proof-of-Work Project Marketplace
Portfolio page: https://priyamjain.me/projects/?id=o6JS8N0ayysiJcn0xb0Q
Scope: Pinchit workspace project. Current automation studio: https://priyamjain.me/work/pinchit/
Category: AI Systems · Marketplace Infrastructure · Reputation Systems · Web Platform
Summary: Pinchit is a shared workspace where people and AI agents work together on projects.
Instead of using agents as isolated tools, teams can bring them into the same workspace, assign them tasks, give them context, and track what they are doing alongside human contributors.
I designed and built the product from its early concept through the first working versions, covering product flows, agent orchestration, workspace architecture, backend systems, and integrations.
Problem: Most AI agents still work outside the team’s normal workflow.
They receive one-off prompts, lose context between tasks, and operate without clear ownership. Teams then have to move information manually between chat, project-management tools, code repositories, and agent interfaces.
The challenge is not access to capable models. It is giving agents enough context, structure, and visibility to participate in real work.
Solution: ## What I Built
* Designed the product and technical architecture for shared human–agent workspaces
* Built systems for workspaces, channels, tasks, ownership, and conversations
* Designed agents as visible workspace members rather than background tools
* Created workflows for assigning work to either humans or agents and tracking execution
* Built persistent context flows across planning, implementation, and execution history
* Added transparent logs showing:
* What an agent was asked to do
* What context it received
* What it produced
* Explored integrations with coding agents such as Claude Code and Codex
* Built Slack-based workflows so teams could work with agents inside existing communication patterns
## Product Flow
```text
Project context
↓
Discussion and task creation
↓
Human or agent ownership
↓
Planning and execution
↓
Visible progress and output logs
↓
Review, feedback, and handoff
```
## Current Direction
Pinchit is evolving into infrastructure for teams that work with multiple AI agents.
The focus is on giving agents:
* Clear ownership
* Persistent project context
* A visible place inside the team’s workflow
* Traceable progress and outputs
* Continuity across planning, execution, review, and handoff
Instead of repeatedly prompting separate tools, teams should be able to assign work, follow progress, review outputs, and continue from the same shared context.
Demo: https://pinchit.in
Image: https://bridgeapp.ai/_next/image?url=https%3A%2F%2Fvaluable-card-aee9b5fcb4.media.strapiapp.com%2Fworkspace_1_da95208901.jpeg&w=1920&q=75
Technology: typescript, NestJs, NextJsNestJS · React · PostgreSQL · Redis · LangChain · LLM APIs · Docker · Auth · Queue Workers
Tags: Marketplace · Proof-of-Work · Reputation · AI Agents · Collaboration · Startup Infrastructure
## Reputation Engine — Proof-of-Work Scoring System
Portfolio page: https://priyamjain.me/projects/?id=fpQGFZg4T0XEDU0mPJic
Category: Trust Infrastructure · Scoring Systems · Marketplace Algorithms · Data Platforms
Summary: A reputation engine that measures contributor quality based on real work instead of resumes or endorsements. It scores impact, reliability, collaboration behavior, and consistency across projects to produce a verifiable proof-of-work reputation graph for marketplaces and collaborative platforms.
Problem: Online marketplaces rely on:
Star ratings.
Self-reported profiles.
Social proof.
These are shallow, gameable, and poor at representing real capability or reliability.
Solution: I designed a system that:
Tracks real contribution events.
Measures task difficulty, quality, and delivery.
Weighs collaboration behavior.
Builds a reputation graph over time.
Outputs verifiable, portable reputation signals.
It turns work history into a living trust network.
Image: https://images.ctfassets.net/az51l532ei6u/7b0EQzZeL0Mc9eoS8xBqB4/11d5d51744603ce6dd529a946e6d3a58/Copy_of_Ult_Rep_Guide-7_1.png?w=3032&h=2952&q=90&fm=png
Technology: Node.js · Python · Graph Databases · Scoring Pipelines · Event Processing · Analytics Stack
Tags: Reputation Systems · Graph Algorithms · Proof-of-Work · Scoring Models · Trust Engineering
## Self Improving agents
Portfolio page: https://priyamjain.me/projects/?id=SgTu65GD8uF9EMx56qVa
Category: AI
Summary: These are self improving agents specially built around social media content creation, they understand your habits find trends and generate content for you , its best to use them longer as it understands your prefrences and get better with time. You need Claude/codex to run it better.
Problem: It makes it much faster to generate content and scripts.
Solution: You get tailor made content and scripts that is currently relevant.
Source code: https://github.com/priyam-jain-2002/auto-agent-creator
Technology: Claude, Codex
Tags: AI
## TruthSeek
Portfolio page: https://priyamjain.me/work/truthseek/
Status: Closed startup · historical case study
Summary: Helped launch an AI voice interview product that reached an enterprise pilot and earned revenue before closing.
Role: Helped build the product from concept through an enterprise pilot, working across the application and voice interview workflow.
Architecture: The Next.js application and NestJS backend connected study setup and review to the Vapi voice integration. The stack used Supabase and OpenAI. Recordings and transcripts gave researchers material to review alongside the analysis.
Engineering impact: Connected study setup, voice interviews and post-interview review into a working customer workflow.
Business impact: TruthSeek served customers, reached an enterprise pilot and generated revenue during its operating period. The startup has since closed.
Workflow: Define a research brief, prepare questions and an audience, run interviews, then review recordings, transcripts and analysis.
Scope: The linked product page documents the historical workflow.
Historical product walkthrough: https://truthseek.in/