shishir lohar

Things I've built.

Applied AI, useful software, and the systems underneath.

In development · 202601

KTPI

An agentic incident investigator.

Using agentic AI with Model Context Protocol (MCP) tools to correlate deployments, error logs, and service health, producing evidence-backed root-cause hypotheses. Built with a 3-person engineering team at Kappa Theta Pi.

  • React
  • PostgreSQL
  • AWS SQS
  • MCP
  • Docker
Read about the work
UGA Warnell · 202602

Forestry AI assistant

A question. A query. A clearer picture.

Local LLMs turn natural-language questions into validated SQL and interactive geospatial maps. Processes 500K+ forestry records, with workflows completing in under 30 seconds.

  • Python
  • Qwen2.5
  • Pandas
  • Streamlit
  • Folium
Read about the work
Founder · 202603

Spotr

Vehicle sourcing, connected.

An AI-powered platform combining 3+ automotive data APIs into one search experience for dealerships. Secured $7,500+ in funding and developed ranking across 10+ criteria.

  • APIs
  • AI
  • Data pipelines
Visit Spotr
Personal project · 202504

Health trend model

Finding patterns in everyday health.

A Random Forest pipeline trained on 1,000 synthetic health records and eight biometric and lifestyle features. Achieved 98% accuracy on a held-out synthetic test set; integrated real-time inference into Streamlit.

  • Python
  • scikit-learn
  • Pandas
  • Streamlit
Visit my GitHub
Forage simulation · 202605

Transaction processing

From events to validated records.

A Spring Boot service built for the JPMorgan Chase & Co. Forage experience, with Kafka transaction processing, SQL persistence, and a REST API.

  • Java
  • Spring Boot
  • Kafka
  • SQL
View experience

More code on GitHub ↗

Go somewhere.

About 01Projects 02Experience 03Open résumé Send an email GitHub
↑ ↓ navigate ↵ open