Skip to content
View Akkikens's full-sized avatar

Block or report Akkikens

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Akkikens/README.md

Typing SVG

Hey there, I'm Akshay πŸ‘‹

I'm an AI Full Stack Engineer based in San Francisco, building things at the intersection of infrastructure, AI, and product. I believe great software is shipped fast, measured obsessively, and built with empathy. That's been my north star across startups, research labs, and enterprise β€” from genomics platforms to multi-agent RL systems to AI career coaches.

Right now, I'm shipping code at Softmax and Climb Together β€” and I've been loving every commit. πŸš€

Β  Β 


🍿 What I'm building right now

Goldi β€” AI Career Coaching Platform @ Climb Together

Goldi is an AI-powered career coaching experience, and I'm the #1 contributor. Think of it as a coach that actually listens β€” with voice mode powered by Hume AI, emotional intelligence built in, and smart interruption handling. We partnered with Walmart to build custom onboarding flows targeting different career archetypes (Explorer, Directed, Networker), and shipped a WhatsApp + SMS consent & accountability system that drove +28% re-engagement. The playlist architecture is URL-driven, multi-playlist, with race condition fixes and navigation caching that just works.

Stack: Next.js Β· Claude AI Β· Hume AI Β· WhatsApp/SMS integrations

Metta β€” Open-Source Multi-Agent RL Platform @ Softmax

Metta is an open-source research platform for multi-agent reinforcement learning, and I own the infra & platform layer. I built a full Datadog + OpenTelemetry observability pipeline β€” metric emission, OTLP receivers, structured collectors that replaced a legacy SQLite setup. On the Kubernetes side, I wrote a watcher service for job lifecycle event processing and storage, plus Helm deploy fixes. Infrastructure-as-code lives in Terraform (Lambda, Secrets Manager, IAM for GitHub webhooks). I hardened the CI/CD pipeline with cancelled-run cleanup, failure classification (timeout / OOM / policy_error), and keep-main-green automation. And I set up Alembic migrations over PostgreSQL + SQLModel ORM for a 21-table schema.

Stack: Python Β· Kubernetes Β· Terraform Β· AWS Β· Datadog Β· OpenTelemetry Β· PostgreSQL


🎯 Impact by the numbers

Here's what shipping fast and measuring everything actually looks like:

⚑ -50%
Lambda cold starts
βœ… 99.9%
Service uptime
🎨 -40%
First Contentful Paint
🏎️ -55%
API latency
πŸ“± +28%
Re-engagement
πŸ§ͺ 500+
Tests written

πŸ”§ Tech I work with



🦦 Featured work

🧠 Metta
Open-source multi-agent RL platform @ Softmax. Built the observability pipeline, CI/CD hardening, Terraform infra, and DB migrations.
🎯 Goldi
#1 contributor. AI career coaching with voice mode, Walmart partnership, playlist architecture. Next.js + Claude AI.
🧬 Factorbook 2.0
Genomics AI platform @ UMass Chan. Led the frontend rebuild β€” cut FCP by 40%, wrote 500+ tests, slashed API latency 55%.

πŸ“Š GitHub stats


πŸ‘“ A bit more about me

AI Full Stack Engineer β€” Softmax Β· Climb Together
Previously @ UMass Chan Medical School Β· Capgemini (BMW) Β· tag8 Β· KPMG
M.S. Computer Science, Clark University Β· AWS Certified Developer Β· San Francisco, CA πŸŒ‰

"Tell me more, I can't get enough!" β€” check out my portfolio or just say hi. I'm always down to chat about infra, AI, or why observability is underrated. β˜•

Pinned Loading

  1. akshay-portfolio akshay-portfolio Public

    TypeScript 1

  2. anthropics/claude-code anthropics/claude-code Public

    Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflo…

    Shell 109k 17.9k