agentic systems for real-world workflows

I’m Lohit Geddam, a software engineer building AI agents, evaluation workflows, and scalable infrastructure for data-heavy systems

status: building the next thing

email linkedin github download resume

about

I build agentic systems that make complex processes easier to run: evaluation loops, data-heavy workflows, multi-tenant FastAPI and React applications, and infrastructure that can survive production constraints

experience

09/25 - 05/26

Technical Co-Founder & CTO

Tandem / San Francisco, CA

Built an AI platform that ties together mechanical engineering context, requirements, and implementation workflows
Built production AI agent workflows with tool calling, semantic search, CRUD operations, and persistent multi-turn context
Authored AWS infrastructure in Terraform across commercial and GovCloud environments

09/24 - 09/25

Software Development Engineer

AWS SageMaker HyperPod / Bellevue, WA

Engineered scalable ML infrastructure for 10k+ GPU superclusters orchestrated by Kubernetes
Built high-concurrency Go features for distributed ML workloads
Reduced node transfer time by 83% and helped reduce MTTR with image pre-caching work

06/23 - 09/23

Software Development Engineer Intern

Amazon Alexa Kitchen / Sunnyvale, CA

Built an Alexa Mobile App feature in TypeScript and React Native for personalized recipe discovery
Implemented a Flask and Python backend that reduced recipe retrieval latency by 80%

06/22 - 09/22

Software Development Engineer Intern

AWS Data Platform / Seattle, WA

Automated PySpark jobs for cleaning and profiling 600GB+ datasets with AWS Glue
Built a React dashboard for customer access to metrics across 1,000+ tables

projects

Percy

Python, SQLite, autonomous agents, option spreads, risk controls

Designed an autonomous stock and options trading system to rank trade candidates, verify decisions with separate model roles, manage option-spread positions, gate execution through deterministic risk controls, and prepare broker-ready order payloads with full SQLite audit history

Fake News Detection

Google Cloud, Vertex AI, Python, LLMs, Generative AI, NLP

Built predictive multi-class models using 25,000+ articles, deployed real-time models with Vertex AI, and used Gemini with RAG and vector databases for feature engineering and fact checking

hobbies

motorcycling weekend rides, backroads, and anything with a good line through a corner
poker play the man, not the cards; poker vlogging at Lucky Lion Poker
basketball find me here Fridays at 7pm

skills

Python Go TypeScript React FastAPI PostgreSQL pgvector Kubernetes Docker Terraform AWS ML infrastructure AI agents

BS Data Science, Minor in Marketing — University of California, San Diego