Headline

Engineering Agentic Infra at WebZero | Ex-MLE @ Sol Foundry Inc. | Building Multi-Agent AI Solutions

About

AI Engineer | LLM Architect | Problem Solver "If you can't explain it simply, you don't understand it well enough..." An AI researcher turned engineer driven by curiosity, creativity, and the pursuit of building useful, scalable systems that work in the real world. Throughout my career, I’ve had the privilege of working with startups and global teams, leading projects in LLM systems, RAG architectures, medical AI, and real-time graph-based reasoning. From creating synthetic emotional datasets and deploying production-grade AI at Vibe AI, to architecting a novel hybrid GraphRAG system at Maura.ai, I specialize in pushing ideas from whiteboard to production fast. At night, I bring out my enigmous side of building quirky projects, whether it's a GAN that generates abstract art, or a voice assistant for HuggingFace documentation, or an open source contributor at Cognee! I’m always open to connecting with people solving hard problems in applied AI, healthtech, and product-driven research!

Location

Bengaluru

AI stack

claude code

cursor

hugging face

langsmith

openai

Goose

Work experience

Founding Engineer at Webzero, Apr 2026 - Present · 5 mo, Bengaluru, Karnataka, India

Machine Learning Engineer at Sol Foundry Inc., Jan 2026 - Mar 2026 · 2 mo, Bengaluru, Karnataka, India, - Architected and built Cerebro, the multi-agent AI runtime service powering the Klaire platform — comprising 10+ specialized LLM agents for message classification, callout extraction, enrichment, duplicate detection, and lifecycle management. - Designed a sequential context-grounded callout extraction pipeline (v3) with entity relationship graphs, ephemeral context stores, and entity resolution — eliminating output fluctuation from batch processing. - Built real-time enrichment and batch onboarding agent flows processing Slack/Gmail messages through multi-stage LLM pipelines with Cerebras inference, LangSmith tracing, and dynamic prompt management. - Built a production evaluation framework with 27 graders across 5 agent types, automated Slack alerting on failures, daily quality dashboards, CI quality gates, and auto-collection of failed traces as golden test cases. - Designed and implemented an end-to-end SLM fine-tuning pipeline (QLoRA on LiquidAI LFM2.5-1.2B) to distill GPT-5-mini into a self-hosted model, reducing classification latency from 2.1s to <200ms and inference costs by ~15x. - Fine-tuned DistilBERT classifiers achieving 87.2% accuracy for multilingual message actionability scoring (English, Hindi, Kannada, Telugu, Tamil, Spanish). - Engineered topic identification agents with content-based deduplication via fingerprinting, resolving cross-topic pollution and duplicate callout issues in production.

Founding AI Lead at Vibe AI, Jan 2025 - Dec 2025 · 1 yr, Bengaluru, Karnataka, India, - Trained and developed the world’s 15th best AI conversational model on EQBench3. - Led and managed the research and engineering team of 3 founding engineers and psychologists. - Planned and researched 4 quantitative and qualitative benchmarks for making the model more human-like. - Developed v1 and v2 algorithms of the conversational agent and engineered the external agents. - Building the foundation behavioral model for user personalization.

AI Engineer at Maura, Aug 2024 - Jan 2025 · 6 mos, San Francisco, California, United States, - Created and implemented a novel architecture for real time graphRAG with contextual updation using OpenAI GPT models and Neo4j for project management domain. - Integrated 10+ core functionalities of the AI like vision, and added slack API support for AI bot creation.

Senior AI Engineer at HealTether Health, Jan 2024 - Aug 2024 · 8 mos, Chennai, Tamil Nadu, India, • Created the AI backend and Hybrid RAG architecture for the doctor’s application platform that performed differential diagnoses detection and associated symptoms detection from patient data along with clinical investigations. • Developed one of India's novel disease-symptom mapping graphs in Neo4j by parsing patient records collected from 20+ hospitals. • Assisted the company to become a partner of NVIDIA Inception Program and represented at NVIDIA AI SUMMIT 2024 at Mumbai. • Led a team of 5 AI engineers to finetune the clinical model and develop the AI backend in FastAPI and Python.

Education

Bachelor of Technology - BTech, Computer Science and Engineering (Data Science), The neotia University

Public links

github: https://github.com/AniLeo-01

cogniqa.space: https://www.cogniqa.space/