UTKARSH TRIPATHI
Projects & Research About

Hello!
I'm Utkarsh Tripathi
AI

I make AI
show its work.

LLM output that can be checked: every claim traced to its source, measured for grounding and unsupported statements, and logged so a reviewer or auditor can follow the reasoning. Built for regulated financial workflows, where a confident wrong answer is the expensive failure.

Founder & CTO of MantaSol, a BFSI tech-compliance company focused on the safe use of AI agents and LLMs in financial services. Before that: data engineering and ML across banking, insurance and edge systems — pipelines, streaming, and analytics at scale. I favour well-defined repeatable workflows over open-ended agents, because auditability depends on the path being specified rather than improvised.

Projects

End-to-end data systems built for scale; from real-time pipelines to ML-powered applications.

Research & Insight

Deep dives into algorithms, statistics, and ML theory; implemented from first principles.

Expertise

Core

  • Python
  • SQL
  • Apache Spark
  • Kafka
  • Flink
  • Airflow

Cloud & Infrastructure

  • AWS (S3, Redshift, Glue)
  • Docker
  • Kubernetes
  • Terraform
  • Git & CI/CD

Data & Analytics

  • Data Modeling
  • dbt
  • Vector DBs (Qdrant)
  • Redis
  • Power BI

AI & LLM

  • LangChain
  • Sentence-Transformers
  • FastAPI
  • LlamaIndex

Evaluation & Assurance

  • RAGAS Evaluation
  • Claim-Level Grounding
  • Unsupported-Claim Rate
  • Hallucination Auditing
  • Adversarial / Prompt-Injection Testing
  • Traceability & Provenance

Experience

Building data and AI systems across industries; from insurtech analytics to verifiable AI for regulated finance.

2021 - 2023

System Engineer

Siemens Ltd.

ETL • Python • Data Pipelines

Jul 2024 - Dec 2024

CV Engineer

Habitat

Edge Inference • CI/CD

Jan 2025 - Sept 2025

Data Engineer

RiskPe

PySpark • SQL • Power BI

Oct 2025 - Present

Founder & CTO

MantaSol

BFSI AI Compliance • RAG Evaluation