Nirbhay Rathod

Proof of Experience

Senior Software Engineer - AI

Publicis Groupe · Jul 2026 – Present

Work directly with the Chief Solutions Officer to shape AI-system architecture, defining agent workflows, specialist responsibilities, routing, handoffs, shared state, and human approval points.
Reduced tokens per query by 30% and lowered hallucination rates from 20–25% to under 10% through context engineering; implemented caching to shorten response times for similar queries.
Use human-aligned LLM-as-a-judge evaluations to assess agent retrieval, answer accuracy, and faithfulness; incorporate reranking and evaluation feedback into iterative improvements.
Implement SQL validation and input/output guardrails targeting prompt injection and harmful content, with LangSmith observability to inspect agent execution and investigate behavior.
Develop MCP servers and agent skills to standardize tool access across agents, simplify third-party integrations, and make specialist capabilities reusable across agentic stacks.
Expanded vector-backed agent memory from semantic memory to include episodic and procedural memory.

Software Engineer (AI Engineering)

Publicis Groupe · Jan 2025 – Jun 2026

Worked directly with the Chief Solutions Officer to turn business briefs into AI products for audience planning, media analytics, social intelligence, and generative content.
Built a LangChain and LangGraph multi-agent system with specialist agents for planning, data retrieval, answer synthesis, and audience sizing.
Implemented semantic agent memory using vector databases for knowledge retrieval.
Developed a multi-provider audience-sizing agent that saved teams approximately 3-5 hours of estimation effort, helping strategists compare audience options and develop better-informed plans.
Implemented table-augmented generation (TAG) / text-to-SQL workflows for querying relational databases, extending AI-assisted information access to structured business datasets.
Built retrieval-augmented generation (RAG) pipelines for textual sources such as trend reports, covering ingestion and retrieval of information for use in AI applications.
Created a conversational analytics assistant for structured TV and digital performance data, enabling users to explore media performance and obtain insights through natural-language questions.
Built social-content intelligence across Instagram, TikTok, and YouTube with scheduled archiving of brand, competitor, and influencer posts, database ingestion, and a ReAct agent for querying the archive.
Created a weekly newsletter agent with tools to fetch and process trends from Facebook, Instagram, X, TikTok, and Google Trends.
Built embedding-based clustering to detect related trends and emerging subcultures across platforms.
Created an AI video-generation platform with scene-level controls and automated stitching, saving production time and enabling creative and non-creative teams to create longer videos without specialist editing skills.
Partnered with TikTok to integrate four Symphony capabilities into the platform: image-to-video generation, caption removal, avatar video generation, and dubbing.
Built rapid Streamlit prototypes to validate implementation feasibility and refine AI capabilities against stakeholder requirements.
Deployed AI applications on AWS Fargate and used Amazon S3 for object storage.

Proof-of-Work

No completed sessions yet. Anything Nirbhay finishes shows up here with the evidence behind it.

Proof of Education

Bachelor of Technology, Metallurgical and Materials Engineering

Indian Institute of Technology Roorkee · 2020 — 2024

Skills