SYSTEMS NOMINAL — OPEN TO AI AGENT & FORWARD-DEPLOYED ROLES
AI Agent Engineer
I build multi-agent systems, agent harnesses, and MCP servers — plus the evals, observability, and infrastructure that make agents reliable in production.
- LOCATION
- San Francisco, CA
- FOCUS
- Multi-agent systems · MCP · Evals
- CURRENT
- AI Engineer @ Climb Together
- SCALE
- 100K+ users in production
- UPTIME
- 99.9%
- STATUS
- Open to agent & forward-deployed roles
TRACE 001 — profile
EST. 2020
Engineer, forward-deployed
Hello! I'm Akshay, an AI agent engineer with 4+ years of full-stack & platform experience building scalable, real-time, production AI systems. I hold an M.S. in Computer Science from Clark University (GPA 3.8, May 2025) and a B.E. in Information Technology from the University of Mumbai (GPA 3.6).
Most recently I worked as an AI Platform Engineer at Softmax — building Claude Code plugins, MCP tool harnesses, and the observability and Kubernetes infrastructure behind a multi-agent RL training platform — while at Climb Together I drive architecture for an AI-powered onboarding platform serving 100K+ users.
I thrive forward-deployed: embedded with customers in fast-paced environments, owning problems from infrastructure to UI and shipping production systems on the ground.
Mission
Looking for AI agent, platform, and forward-deployed engineering roles where I can build multi-agent systems, harnesses, evals, and observability that ship production AI — embedded with customers, with high ownership.
AGENTS & LLM
- Claude Code & MCP
- Multi-Agent Orchestration
- Evals & Tool Use
- Anthropic / OpenAI APIs
- LangChain
- Hume AI
- Multi-Agent RL
- Prompt & Context Engineering
LANGUAGES
- TypeScript / JavaScript
- Python
- Java
- C#
- SQL
- C++
PRODUCT & BACKEND
- Next.js / React
- Node.js
- Django
- Spring Boot
- GraphQL / REST
- PostgreSQL (Drizzle ORM)
PLATFORM & INFRA
- AWS (Lambda, S3, EC2)
- GCP
- Kubernetes & Helm
- Terraform
- Docker
- GitHub Actions CI/CD
- Datadog
- OpenTelemetry
- users scaled
- 100K+
- OSS commits
- 1,180+
- Claude Code skills
- 86
- tests written
- 500+
- production uptime
- 99.9%
- cold-start latency
- −50%
TRACE 002 — showreel
Agents that think.
MULTI-AGENT ORCHESTRATION · TOOL USE
Shipped to production.
HARNESSES · MCP SERVERS · EVALS
Reliable at scale.
OBSERVABILITY · KUBERNETES · 99.9% UPTIME
TRACE 003 — experience
2020 — PRESENT
Where I've worked
Climb Together
May 2025 — PresentFull Stack Developer & AI Engineer San Francisco, CA
- Built an AI-powered chatbot with LangChain, OpenAI, and Anthropic Claude; added voice via Hume AI for natural conversations.
- Drove architecture for a distributed onboarding system (Next.js, Clerk, Twilio) scaling to 100K+ users with a 35% lift in verified opt-ins.
- Built real-time data pipelines powering SMS nudges and behavior analytics, driving a 28% increase in feature re-engagement.
- Drove migration to serverless (AWS Lambda) and PostgreSQL with Drizzle ORM, cutting cold-start latency by 50%.
- Set up CI/CD with automated testing, linting, and environment-specific deploys to Vercel; instrumented Sentry error tracking and PostHog product analytics.
- Led architectural reviews enforcing modular design, code ownership, and performance benchmarks.
Softmax
Nov 2025 — Apr 2026AI Platform Engineer (Contract) San Francisco, CA
- Built and maintained Claude Code plugins (86 skills across 6 plugins) and MCP tool harnesses powering AI-assisted dev workflows — autonomous PR review, code generation, and multi-file editing across the org.
- Designed a production observability pipeline (Datadog + OpenTelemetry: OTLP receivers, structured collectors), replacing legacy monitoring for a multi-agent RL training platform.
- Built a Kubernetes watcher service for AI job lifecycle orchestration — automated failure classification (timeout/OOM/policy_error), stuck-job reconciliation, and keep-main-green CI.
- Engineered Terraform IaC for AWS (Lambda, Secrets Manager, IAM) powering GitHub webhook integrations and deploy pipelines.
- Built a tournament system with daily reporting cronjobs, error-log surfacing, and diagnostic scoring to evaluate multi-agent RL policy performance.
- Owned the data layer: 21-table PostgreSQL schema with Alembic migrations and SQLModel ORM; pre-deploy Helm migration jobs for zero-downtime releases.
- Shipped 1,180+ commits to the Metta AI open-source multi-agent RL platform.
UMass Chan
May 2024 — Jan 2025Software Engineer Intern Boston, MA
- Led implementation of Factorbook 2.0 (React, Next.js, GraphQL), improving First Contentful Paint by 40%.
- Integrated real-time dashboards with Prometheus to monitor distributed job queues and system health.
- Wrote 500+ unit and integration tests with Jest and Playwright, maintaining 98%+ coverage.
- Ran sprints, standups, and retros for a 6-person agile pod; built data-heavy genomic visualizations.
Capgemini
Aug 2021 — Aug 2023Full Stack Engineer Mumbai, India
- Built 10+ microservices (Java, Spring Boot) for BMW's global platform — 200K+ users, 99.9% uptime, and a 55% performance gain.
- Built e-commerce modules with React, Redux, and Webpack, cutting page load times by 50%.
- Migrated a legacy monolith to Vue.js, improving performance by 40%.
- Owned CI/CD with Jenkins and Docker, reducing manual deploy effort by 70%; mentored 4 junior developers.
tag8
Aug 2020 — Aug 2021Software Engineer Mumbai, India
- Integrated Redis caching and Elasticsearch with a Django backend, cutting database query latency by 60% and powering real-time dashboards.
- Lifted customer satisfaction by 25% through a UI/UX redesign of high-traffic flows.
KPMG
Jul 2020 — Dec 2020Software Engineer Intern Mumbai, India
- Shipped 30+ CRM feature enhancements in Java; contributed to distributed application architecture and data-handling compliance.
TRACE 004 — projects
SELECTED WORK
Things I've built
DevDiagrams — Interactive Learning Platform
Interactive learning platform built with Next.js 15 and React, featuring comprehensive UI components from Radix UI. Integrated Supabase for authentication and real-time data management, with Stripe and Razorpay for payment processing. Includes PDF processing, AI-powered features using OpenAI, and real-time notifications — optimized with Vercel Speed Insights.
AI Chatbot Platform
Advanced AI chat application built with Next.js, React, and Supabase. LangChain conversational AI, Hugging Face models for specialized NLP tasks, and OpenAI integration for intelligent responses — with real-time analytics, conversation memory, and multi-agent orchestration.
Advanced ML Pipeline & Agent System
End-to-end ML pipeline with multi-agent orchestration: data preprocessing, PyTorch/TensorFlow training, evaluation pipelines, and deployment automation. Agent-based systems for automated data processing, model validation, and intelligent decision-making.
AI-Powered Career Platform
Intelligent web application connecting Clark University students with alumni for career guidance — LangChain matching, Hugging Face resume analysis, and OpenAI-personalized recommendations. Reduced manual effort by 80%.
TRACE 005 — open-source
1,180+ COMMITS
Contributions in production
Softmax
Metta AI — Multi-Agent RL Platform
1,180+ commits to an open-source multi-agent reinforcement-learning platform: Claude Code plugins (86 skills across 6 plugins), MCP tool harnesses, a Kubernetes job-lifecycle watcher with automated failure classification, a Datadog + OpenTelemetry observability pipeline, and the 21-table PostgreSQL data layer with zero-downtime Helm migrations.
Climb Together
Goldi — AI Career Assistant
Core engineer on an AI career assistant that teaches students a better way to land internships and jobs. Conversational AI with LangChain, OpenAI, and Anthropic Claude; natural voice via Hume AI; distributed onboarding architecture that scaled to 100K+ users. Supported by Google.org and Walmart.org.
UMass Chan Medical School
Factorbook 2.0 — Genomic Research Platform
Led implementation of a genomic research platform (React, Next.js, GraphQL) — 40% faster First Contentful Paint, data-heavy genomic visualizations, Prometheus dashboards for distributed job queues, and 500+ tests at 98%+ coverage.
TRACE 006 — credentials
VERIFIED
Certifications
TRACE 007 — signals
PEER REVIEW
What people say
As the main UI developer on the project, Akshay undertook a very comprehensive overhaul of one of our resources, Factorbook, and modernized it using Next.js and MUI. He completed the redesign of Factorbook fully from the ground up, and was responsible for all parts of the frontend of the application. Very impressive and complex project. He was a team player and was always very receptive to feedback and looking for ways to improve the site and his own skills.
Having worked with Akshay for 2 years, I must say he is an excellent professional. I am impressed by his work ethics, communication skills, coding skills, exceptional problem-solving skills. He truly has the skills to interact with business clients. But what makes him stand out is his willingness to help others. I have no hesitation recommending him to potential employers.
Great colleague who is very open minded when it comes to both personal and professional development in the software engineering industry. Akshay is always going above and beyond, ready to tackle new and difficult tasks, and ready to take responsibility. Very fast onboarding to the team, great attention to detail and a critical eye to improve things make him a great addition to any team!
I studied with Akshay at the University of Mumbai, and he was always someone you could count on — smart, focused, and a great team player. He's technically strong and quick to pick up new concepts, and I'm confident he'll be a great asset wherever he works next.
TRACE 008 — handoff
Ready to ship agents that work?
I'm open to AI agent, platform, and forward-deployed engineering roles — or a conversation about multi-agent systems, MCP servers, evals, and shipping production AI. Reach out and I'll get back to you quickly.