Karun Pabbi
Machine Learning and GenAI Expert
I help organizations ship AI that works in scalable production systems.
Seven years building GenAI, LLM and ML systems at Capital One, JP Morgan, EXL and Thomson Reuters: LLM fine-tuning, evals, token optimization, RAG, agentic workflows and custom harness development. Book a session for career guidance, an AI build, your startup’s MVP and go-to-market, or a standing advisory role.
What the work changed
Measured results from systems I designed and led.
Model documentation, time to market
18+ months~30 daysModel revalidation
7 days4–5 hoursSupport resolution time
Baseline72% fasterML platform downtime
Baseline43% less
Built systems at
- Capital One
- JP Morgan
- EXL
- Thomson Reuters
- Infosys
- Turing - Gemini
- Turing - Apple Intelligence
What I build
The parts of a GenAI system that decide whether it survives production: quality you can measure, cost you can predict, and failures you can trace.
GenAI and LLM systems
RAG, agentic workflows and LLM applications built for enterprise scale and regulated data.
LLM fine-tuning
Supervised fine-tuning and parameter-efficient methods (LoRA, QLoRA) on open models like LLaMA, when prompting is not enough.
Evals
Offline eval suites, LLM-as-judge scoring, regression gates in CI and production quality monitoring.
Token optimization
Prompt compression, caching, model routing and context design that cut latency and cost per call.
Harness engineering
The scaffolding around a model: tools, memory, retries, guardrails and observability that make agents reliable.
Custom harness development
Purpose-built AI harnesses for triage, root cause analysis and model governance, like the one that cut support resolution time by 72%.
Work with me
Every session is booked and paid here. Pick a time in your own time zone; you get a calendar invite and a video link straight away.
Career guidance
Resume review, role targeting and interview prep for data science, ML and GenAI roles.
60 min call, or a 3-session pack
Price on requestBookAI and ML implementation
Scope, architect and de-risk a GenAI or ML system: fine-tuning, evals, token cost and a custom harness for production.
90 min deep dive, or a project engagement
Price on requestBookStartup MVP and GTM
Turn an idea into a testable MVP and a go-to-market plan with clear first customers.
2-hour working session, or a 4-week sprint
Price on requestBookAdvisor onboarding
A standing advisor for your AI, data or fintech startup: monthly calls and async access.
Starts with a free 15 min intro call
Price on requestRequest intro
The first posts on GenAI, startups and fintech are on their way.
About
I build machine learning systems banks can trust in production: LLM agents that write and validate model documentation, RAG tools that find the root cause of platform failures, and anomaly detection that catches errors before traders do.
MTech in Data Science and Engineering, BITS Pilani. 4th of 50,000 in the Infosys Global Innovation Hackathon. Founder of Saanjisath; helps run Punjabi in Bengaluru.
More about my workGet in touch
Not sure which service fits? Send a note and I’ll reply within two working days.