B.Tech CSE (AI & Data Science) — IIIT Kottayam, Class of 2027
Microsoft Certified: Azure AI Engineer Associate
Certified Meshery Contributor (CMC) — Layer5 / CNCF
LFX Mentorship 2026 Applicant — Meshery AI Adapter + Agentic CI Pipelines
Competitive Programmer — 350+ LeetCode problems solved
Hackathon Builder — Varroc Eureka 3.0 & Smart India Hackathon
Powerlifting Champion — IIIT-K Sports Meet 2026

I build end-to-end AI systems — from multi-LLM inference backends and real-time computer vision pipelines to full-stack dashboards and cloud-native tooling. Currently contributing to the CNCF Meshery ecosystem and applying for LFX Mentorship 2026 Term 2.

Check out my full portfolio: portfolio-akshay-sriram.vercel.app

Live Demo: aeo-diag-akshay.vercel.app  |  Python · FastAPI · Groq · Cohere · Ollama · React · Render · Vercel

A production multi-LLM backend with two modes — AI Engine Optimization scoring and a Meshery AI Adapter proof of concept.

AEO Diagnostic:

• Queries Groq (Llama 3.3, Llama 3.1) and Cohere Command-A simultaneously using Python ThreadPoolExecutor
• Sub-3-second multi-provider responses with mention detection, position ranking, and sentiment scoring
• Structured JSON output pipeline with robust response parsing across inconsistent LLM formats

Meshery AI Adapter Prototype (CNCF open source contribution):

• FastAPI endpoint accepts natural language infrastructure intent and returns valid Meshery design manifests
• Output aligned with designs.meshery.io/v1beta1 schema
• BYOM provider switching across Groq, Cohere, and Ollama (local inference) — zero code changes
• Multi-provider parallel mode queries all cloud providers simultaneously and returns best valid schema
• Built as proof of concept for issue #19092

Varroc Eureka 3.0 Hackathon  |  Python · OpenCV · YOLOv8 · FastAPI · React · NumPy

Real-time ADAS system that scores two-wheeler riding behavior 0–100 with live video analysis.

• 60fps WebSocket video streaming with React + FastAPI dashboard
• Constant-Acceleration Kalman filter on ego-motion-compensated centroids via Lucas-Kanade optical flow
• Calibration-free speed estimation using fleet-shared MPP (m/px = w_real / w_bbox)
• 6 behavioral metrics: hard brake, aggressive acceleration, lane weave, tailgating, sudden stop, helmet non-compliance
• Helmet detection via Otsu blob analysis with 30-frame majority vote + Re-ID by proximity
• Written about in depth: Medium blog post

Next.js 14 · TypeScript · PostgreSQL · Prisma · Tailwind CSS · Nodemailer

Full-stack Cal.com-style scheduling app with production-grade architecture.

• Event type CRUD with unique URL slugs & per-day availability rules with timezone support
• Real-time double-booking prevention via half-open interval overlap checks in Prisma serializable transactions
• Normalized 6-table PostgreSQL schema with composite indexes on (event_type_id, date, status) for fast slot queries
• Denormalized end_time for efficient overlap detection
• Automated email notifications for confirmations, cancellations & 24-hour reminders via secure Cron jobs
• 11 REST API endpoints · JWT authentication · Deployed on Vercel

Meshery / Layer5 (CNCF)  |  Active Contributor — 2026–Present

TypeScript · React · Go · GitHub Actions

• PR #1493 approved in Sistent — Layer5's React/TypeScript CNCF design system (added AssignmentTurnedInIcon)
• Certified Meshery Contributor (CMC) — completed all 5 exam units (mesheryctl, Extensibility, Models, Server, UI)
• Reported and documented 2 production bugs in Layer5 Cloud platform:
  • CMC exam submission flow — backend records attempt but frontend shows "Not Attempted"
  • Certificate rendering — Cannot read properties of null (reading 'map') on production-v1.0.73
  • Both acknowledged by @leecalcote on Slack
• Submitted attendance PRs to meshery/meshery for weekly Newcomers meetings
• Active engagement on issues #19092 (AI Adapter), #19012 (CI Pipelines), #19419 (GitHub Actions), #18751, #18753
• Responded to leecalcote's call for Meshery MCP Server volunteer maintainers on the Layer5 developers mailing list
• Applying for LFX CNCF Mentorship 2026 Term 2 — Meshery AI Adapter + Agentic CI Pipelines projects

Estimation of Failure of Airplane Engines  |  Student Researcher — DIAT, Pune (Dr. Yogeshwar Singh)

• Built UI dashboard for data preprocessing and fusion workflows
• Deployed ML models for Remaining Useful Life (RUL) estimation on the NASA C-MAPSS dataset with multi-sensor time-series data

Non-Invasive Detection of Diabetes  |  Research — IIT Bombay (Dr. Mahesh Parihar)

• Trained classification models (Logistic Regression, KNN, Decision Trees) on Pima Indians Diabetes Dataset
• Achieved 96% accuracy with ROC-AUC benchmarking

{
  "languages":     ["C", "C++", "Python", "JavaScript", "TypeScript", "Java", "SQL", "Go"],
  "ai_ml":         ["LLM APIs (Groq, Cohere, OpenAI, Ollama)", "TensorFlow", "Keras", "OpenCV", "YOLOv8", "Scikit-learn"],
  "web":           ["React", "Next.js 14", "FastAPI", "Node.js", "Tailwind CSS", "Prisma", "WebSockets"],
  "databases":     ["PostgreSQL", "MongoDB", "SQL"],
  "devops":        ["Git", "Docker", "GitHub Actions", "Vercel", "Render"],
  "cloud":         ["Microsoft Azure (AI Engineer Associate ✓)"],
  "core_cs":       ["System Design", "OOPS", "OS", "DBMS", "Compiler Design", "Computer Networks"]
}

CNCF Open Source — Certified Meshery Contributor (CMC), PR #1493 merged in Sistent
LFX Mentorship 2026 Applicant — Meshery AI Adapter + Agentic CI Pipelines (CNCF)
Varroc Eureka 3.0 Hackathon — Built real-time ADAS scoring system (National Level)
Smart India Hackathon 2024 — Participant
Powerlifting — Winner, Intra IIIT-K Sports Meet 2026
Basketball — Winners, School-level & Intra IIIT-K Sports Meet 2023
350+ LeetCode problems solved (DSA)
Microsoft Azure AI Engineer Associate — Certified (Valid: July 2025 – July 2026)

Got a project idea, research opportunity, or just want to connect?

Email — akshaysriram.b@gmail.com Medium — medium.com/@akshaysriram.b