AiMi | Full Stack AI Engineer | Remote(Everywhere) | Full-time
AiMi builds agentic AI infrastructure for capital markets: autonomous agents that ingest venue and vendor notifications, assess impact across client infrastructure, and drive change management workflows in real time, replacing what used to be manual, static processes for exchanges and their ops teams. live production platform, not a pilot.
you'd work across the full stack: React/TypeScript frontend, Node.js backend, AWS Lambda. concretely: shipping features in the agent pipeline, building MCP-based connectors and skills that extend what our agents can reach, running evals on agent accuracy and reliability, building observability and alerting for agent latency/error rates/failed workflows, and making long-running LLM, agent, and SSE calls actually resilient (timeouts, retries, circuit breakers, fail-fast, recovery). also a fair amount of diagnosing dependency/version conflicts in a Node.js/npm monorepo, which is its own skill.
looking for 2-3+ years hands-on experience, solid Node.js and REST API background, real hands-on work with LLMs or agent frameworks (Mastra, LangChain, LlamaIndex, Claude, OpenAI or similar), comfortable with AWS (Lambda, Bedrock, DynamoDB, S3, Cognito, API Gateway, CloudWatch), MongoDB/DynamoDB, and genuinely tests what you ship (unit, integration, e2e). familiarity with spec-driven development (Claude Code, Cursor, CodeRabbit or similar) is a real plus since that's how we work day to day. bonus points for capital markets/trading infra background, LLM observability tooling like Langfuse, SSE streaming experience, or PostHog.
you'd be working on great things here: real architectural decisions at an early-stage company, visible impact from week one, and room to grow into a senior or lead role as we grow.
I see you're looking for a Full-Stack AI Engineer with experience in Node.js, AWS, and LLMs. Let's get started.
I'd focus on building robust, resilient agent pipelines and LLM integrations. I'd start by evaluating and improving the agent accuracy and reliability. I'll also build observability and alerting for agent performance. I have experience with LLMs and agent frameworks, and I can help ensure your LLM calls are resilient. I can also assist with diagnosing and resolving dependency and version conflicts in your Node.js monorepo. I'll bring this experience to the table, and I can share more on my approach in our next conversation.
What specific areas of your agent pipeline do you want to improve first?
If you have worked with any Object Relational Mapping tool like Drizzle, Prisma, TypeORM, Django, Rails you have probably come across this problem:
Two developers are working on branches and they both create a migration file named 004.
One of them renames their file to 005 then you merge the changes deploy them and hope that everything works out fine.
Sometimes it does work.. Sometimes a column just disappears because the migration files were run in the wrong order.
We had to deal with this problem many times, including a few times when we had to stay up late at night to fix things that we decided to think of a new way to handle migrations.
Most Object Relational Mapping tools use an numbered system for migrations. This system works fine when one developer is working on a project but it does not work well when there are multiple branches. You end up having to coordinate the migrations dealing with conflicts or worse you have to deal with issues that only show up when the project is live.
Rolling back changes is another problem. Writing rollback scripts is hard and people often skip this step or get it wrong especially when multiple changes are involved.
That is why we built Migratex.
Of using a numbered list migrations in Migratex form a directed acyclic graph, which is similar to how Git models commits.
Each migration file declares what it depends on so multiple branches can create migration files without any conflicts.
When you merge changes:
* If two migration files truly conflict, for example if they try to change the column to different types Migratex flags the problem.
* If the migration files are independent for example if they change tables or columns they merge cleanly.
You do not have to rename any files. You do not have to coordinate with anyone. You do not have to worry about deployments because of ordering issues.
Migratex also generates SQL code automatically:
* You update your Object Relational Mapping schema.
* You run the migratex generate command.
* It compares the schema to the database and generates the migration SQL code.
Migratex also generates rollback scripts alongside the migration scripts to make rollbacks safer.
The goal of Migratex is simple:
Make schema changes safe in an environment with developers, without relying on manual ordering or fragile rollbacks.
I built link-guardian, a Rust command-line tool that scans websites and GitHub repositories to detect broken and redirected links.
It parses HTML and Markdown (including READMEs), checks links concurrently using async networking, and outputs both human-readable and JSON reports. It’s designed to work well in CI pipelines by returning a non-zero exit code when broken links are found.
Motivation: I often found outdated/broken links in docs and READMEs and existing tools were either slow, language-runtime dependent, or not GitHub-aware. Rust made it possible to ship a single fast static binary.
Git Quick is a modern CLI tool that speeds up repetitive Git commands with smart defaults and AI-powered commit messages. It combines git add, commit, and push into a single command while generating meaningful commit messages from your code changes.
Key Features:
- Single command workflow: Just run 'gq' to stage, commit with AI-generated message, and push
- AI-powered commits: Uses Ollama (local, free), OpenAI, or Anthropic to generate contextual commit messages from diffs
- Works immediately: Smart fallback mode requires no setup - AI is optional
- Commit history visualization: 'gq story' shows compact, colorized commit summaries grouped by date, author, or type
- Time tracking: Built-in development time tracking per branch with detailed reports
- Branch sync: 'gq sync' safely updates all local branches with conflict detection
- Cross-platform: Available via pip (Python 3.8+) and npm
The tool is designed for developers who want to maintain good commit hygiene without the friction of writing messages manually. It runs Ollama locally by default for privacy, or you can use cloud AI providers.
AiMi builds agentic AI infrastructure for capital markets: autonomous agents that ingest venue and vendor notifications, assess impact across client infrastructure, and drive change management workflows in real time, replacing what used to be manual, static processes for exchanges and their ops teams. live production platform, not a pilot.
you'd work across the full stack: React/TypeScript frontend, Node.js backend, AWS Lambda. concretely: shipping features in the agent pipeline, building MCP-based connectors and skills that extend what our agents can reach, running evals on agent accuracy and reliability, building observability and alerting for agent latency/error rates/failed workflows, and making long-running LLM, agent, and SSE calls actually resilient (timeouts, retries, circuit breakers, fail-fast, recovery). also a fair amount of diagnosing dependency/version conflicts in a Node.js/npm monorepo, which is its own skill.
looking for 2-3+ years hands-on experience, solid Node.js and REST API background, real hands-on work with LLMs or agent frameworks (Mastra, LangChain, LlamaIndex, Claude, OpenAI or similar), comfortable with AWS (Lambda, Bedrock, DynamoDB, S3, Cognito, API Gateway, CloudWatch), MongoDB/DynamoDB, and genuinely tests what you ship (unit, integration, e2e). familiarity with spec-driven development (Claude Code, Cursor, CodeRabbit or similar) is a real plus since that's how we work day to day. bonus points for capital markets/trading infra background, LLM observability tooling like Langfuse, SSE streaming experience, or PostHog.
you'd be working on great things here: real architectural decisions at an early-stage company, visible impact from week one, and room to grow into a senior or lead role as we grow.
apply: https://join.com/companies/aimitechnology/16658631-agentic-a... , or reach out directly at vishnu.swaroop@aimi.technology
reply