Vansh Talwar

Software Engineer

Delhi, Delhi, India1 mo experience

AI EnabledAI ML Practitioner

Key Highlights

Reduced API latency by 35% at TechAivv.
Built a real-time chat app with Socket.io.
Designed evaluation pipelines for LLMs at Xelron AI.

Stackforce AI infers this person is a Full-Stack Developer with expertise in AI and real-time systems.

Contact

Skills

Core Skills

Large Language Models (llm)Ai

Other Skills

DockerPython (Programming Language)BenchmarkingData PipelinesReinforcement Learning Human Feedback (RLHF)API DevelopmentQuality AssuranceFull-Stack DevelopmentReact.jsNode.jsJavaScriptREST APIsMERN StackArtificial Intelligence (AI)Systems Design

About

Reach me out at vipultalwin99@gmail.com I build full-stack web applications that actually ship from REST APIs to real-time systems. Over the past year, I've reduced API latency by 35% at TechAivv, built a real-time chat app with Socket.io supporting live messaging across multiple simultaneous users, and trained LLMs at Xelron AI by designing evaluation pipelines and benchmark datasets. Stack: React · Node.js · Express · MongoDB · Socket.io · JWT Auth · Docker

Experience

1 mo

Total Experience

1 mo

Average Tenure

1 mo

Current Experience

Ethara ai

Software Engineer

May 2026 – Present · 1 mo · Gurugram, Haryana, India · On-site

Working on LLM evaluation pipelines and AI agent benchmarking datasets and collaborating with Quality Leads and AI Ops teams.
Contributing to RLHF, RLVR and SFT pipelines and coherence across frontier LLM training workflows and Rubics Evals

DockerPython (Programming Language)Large Language Models (LLM)AI

Xelron ai

Software Engineer

Dec 2025 – May 2026 · 5 mos · Delhi, India · Remote

Contributed to two production AI projects for Snorkel AI.
Designed and completed evaluation involving Oracle tests, NOP tests, and models quality checks, consistently maintaining a high task approval rate throughout the project lifecycle
Generated and submitted training for frontier LLM models (including Claude) grounded in real-world GitHub pull request scenarios, achieving a high approval rate across submissions, used AI tools like Cursor.

DockerBenchmarkingAILarge Language Models (LLM)