Home Skills Blog Publications Contact
Karthik Raja Anandan

Karthik Raja Anandan

Founding ML Engineer (Data Scientist) · Elas

Founding ML Engineer (Data Scientist) at Elas (Pleasanton, CA), building and validating production agentic AI systems — test automation, multi-agent routing, and dev infrastructure. M.S. AI (NLP), UC Santa Cruz (2025, GPA 3.9). Open-source: Google Tunix, LangChain, Ivy. NVIDIA GTC 2026 · SemEval-2025 Task 3 Best System Paper (co-author).

Pleasanton, CA kanandan@ucsc.edu

M.S. AI

UC Santa Cruz — 3.9 GPA

4 Roles

Elas · Samsung · Citi · MultiOn

GTC 2026

NVIDIA Demo — Meta Booth

Open Source

Tunix · LangChain · Ivy

M.S. in Artificial Intelligence (Natural Language Processing)

University of California, Santa Cruz

Sep 2024 - Dec 2025 GPA: 3.9/4.0

B.E. in Computer Science

SSN College of Engineering

Aug 2019 - Aug 2023 GPA: 3.9/4.0
🎙️ Voice AI × Evals

CallLens — See Inside Every Voice-Agent Call

Per-turn root cause for Vapi voice calls: latency by stage, mishearing, dead air. TypeSafe Jev judges every turn in ~120 ms; one click self-heals the live agent. Built at the Vapi hackathon.

Voice AIAI AgentsEvalsObservability
🤖 Robotics × LLM Agents

Morph — Self-Evolving Robot-Fleet Harness

LLM agents write and evolve skills for robot fleets; a verifier blocks unsafe edits. MuJoCo sim.

RoboticsAI AgentsPythonMuJoCo
🛠️ Cal Hacks 12.0

MindNest — AI Personas + SO-101 Robot Arm

My first hardware build: AI personas driving a real SO-101 arm (rock-paper-scissors) via a FastAPI API.

RoboticsHardwareAI AgentsHackathon
⚡ <15 ms routing

Smart Router — Multi-Modal LLM Router

Picks the cheapest capable LLM per query in under 15 ms, with guardrails and fallback.

LLM InfraAI SafetyPythonEmbeddings
🧮 4 backends · 3 chips

Kernel Dev — Cross-Stack Kernels

Softmax in CUDA, Triton, JAX and MLX + matmul roofline study on M4, T4 and TPU v5e.

GPU KernelsML SystemsCUDATriton
🎤 NVIDIA GTC 2026

Helion Optimizations

Helion → Triton kernels profiled for occupancy and bandwidth; demoed at NVIDIA GTC 2026.

GPU KernelsML SystemsHelionTriton
🛠️ OpenEnv Hackathon

StaffingGym — RL Environment for LLM Agents

RL environment where an LLM agent runs a staffing agency for a simulated 52-week year.

Reinforcement LearningAI AgentsPythonOpenEnv
🛠️ Nozomio Hackathon

TeamHQ — Autonomous Code-Change Hub

Proposed code changes become cards; a coding agent turns them into pull requests.

AI AgentsDeveloper ToolsTypeScriptCoding Agents

LLM Evaluation Framework

Multi-model eval harness: RAG, voice, tracing and regression detection.

EvalsLLM InfraPythonRAG

Distributed URL Shortener

Go URL shortener: consistent hashing, Redis cache, OpenTelemetry.

BackendDistributed SystemsGoRedis

Real-Time Log Processing Pipeline

Real-time log pipeline: Kafka ingest, Faust/Spark processing, analytics.

Data EngineeringDistributed SystemsKafkaFaust

Sentiment Analysis

machine-learningnlppythonsentiment-analysis

Data Scientist — Founding ML Engineer

Elas

Jan 2026 - Present Pleasanton, CA
100+ DAU · ~1 s TTFT · GRPO post-training · 160-scenario eval suite
PythonAWS ECSKubernetesArgo CI/CDMulti-AgentIntegration Testing

Data Analysis & Machine Learning Intern

Elas

Jun 2025 - Sep 2025 Pleasanton, CA
Internship → full-time return offer
PythonFastAPIPostgreSQLDockerLLM Agents

AI Research Engineer

Samsung Research Capstone

Sep 2025 - Dec 2025 Santa Cruz, CA
1.5-bit 32B Qwen validated · long-horizon agent verification
PythonGUI AgentsvLLMQwenQuantizationRTX 3070

Software Developer (C09)

Citi Corp

Jul 2023 - Sep 2024 Chennai, India
millions of events/day · 25% faster processing
JavaSpring BootKafkaHadoopSparkJenkins

NLP Engineering Intern

MultiOn (AI agents startup)

Jun 2023 - Aug 2023 Remote
LangChain PR #12392 merged · +40% agent accuracy
PythonLangChainVector DatabasesOpen Source

Core Competencies

Backend & Systems

Distributed SystemsMicroservicesREST APIsEvent-DrivenSystem Design

AI / ML

LLM SystemsRAG PipelinesAgent ArchitecturesPyTorchNLP

Cloud & DevOps

AWS (EKS, EC2, S3, SQS)DockerKubernetesCI/CD

Languages

PythonJavaC++GoRustSQL