VP of AI, Trilogy · Founder & CTO, Tandav Labs

Builder.
Applied AI.
AI research.

Sixteen years of building companies, from RFID and IoT, to social analytics at billion-post scale, to applied AI.

16Years building
3Startups co-founded
10Products shipped at Tandav
11Research articles
01

Builder

Co-founded three startups across IoT, analytics and SaaS. Led product and engineering teams of 5 to 30 in Bengaluru, Berlin and the U.S.

02

Applied AI

Ten products shipped through Tandav Labs, including Noah, an agentic colleague for finance teams. At ROQ.ai, led the teams building a platform that generated working web apps from plain-text prompts.

03

AI research

Training specialized, opinionated language models at Trilogy. Published benchmarks on agentic retrieval, reinforcement learning for agents, and LLM-driven strategy optimization.

01

Right now

Focused language model training
Vice President of AI · 2025 to now

Trilogy

Training specialized, opinionated language models for focused tasks. Researching, benchmarking and deploying the latest in AI, and consulting across orgs on AI transformation of products and processes.

Tandav Labs multimodal AI pipeline
Founder & CTO · 2024 to now

Tandav Labs ↗

An AI research and applied AI lab. Focused language models and agentic products, built and shipped by a deliberately small team.

02

Selected Work

Agentic products, applied AI and full applications I have designed, built and shipped through Tandav Labs with a deliberately small team.

03

Writing

Research notes and benchmarks from Trilogy's AI Center of Excellence, published on Substack.

All 11 articles →
  1. Skip the $600 Mac mini. Run OpenClaw securely on a remote box.

    The setup, the gotchas, and three Claude Code skills that do the install for you in 30 minutes.

    ↗
  2. Building the AI COE Chatbot

    Willfully over-engineering a simple RAG bot to explore agentic workflows.

    ↗
  3. Reinforcement Learning for Agents, Part II

    A comparison of Agent Lightning, Handit.ai, and a homegrown tool, AgentEvolve.

    ↗
  4. Reinforcement Learning Techniques to Optimize Agents

    Can RL loops continuously refine prompts, tools, and agentic pipelines?

    ↗
  5. Auto-Improve Bitcoin Algo Trading Strategies with LLMs

    Evolving a naive strategy from a −2.06 to a +3.99 Sharpe with multi-model LLM loops.

    ↗
04

Career

Praveen Koka

Praveen Koka

CTO · CPTO · Software architect

I have founded and built startups across IoT, analytics and SaaS, leading product and engineering teams of 5 to 30 across Bengaluru, Berlin and the U.S.

At Greenroom our analytics engine processed millions of social posts a day. At Decklar we built IoT tracking for global logistics. As VP Engineering at Berlin's ROQ.ai I led the teams building an AI platform that generated working web apps from plain-text prompts.

  • 2025Trilogy · VP of Artificial Intelligence
  • 2024Tandav Labs · Founder & CTO
  • 2021ROQ.ai · VP Engineering · prompt-to-app platform, Berlin
  • 2020Hawtlist · Founder & CEO · social commerce
  • 2015GOIN’ · Technical Lead · Seattle, remote
  • 2015Knotable · Software Engineering Consultant · New York
  • 2015Greenroom · Co-founder & CPTO · creator analytics at scale
  • 2011Decklar · Co-founder · IoT tracking for logistics
  • 2010Skope Solutions · Co-founder & CTO · RFID systems integration

SpeakingDigitale Leute, Cologne 2022 · iJS Conference, Munich 2022