Uday Shankar Gattu
AI Engineer at PraxisPro
Developing therapeutic-area small language models (SLMs) and AI agents for Life Sciences at PraxisPro.
Intro
I'm Uday Shankar Gattu, an AI Engineer currently at PraxisPro where I build therapeutic-area small language models and AI agents for Life Sciences. My journey has taken me from optimizing high-volume data pipelines at TCS to being a Founding AI Engineer at Galexor AI, where I built an adversarial testing platform that secured pre-seed funding. I recently completed my MS in Computer Software Engineering at Northeastern, where I also taught Prompt Engineering and Gen AI. I’m passionate about moving AI from experimental notebooks to production-grade APIs, with a ruthless focus on latency and reliability. I love discussing agentic workflows, multi-agent orchestration, and AI security. Whether it's shipping an MVP or refining stateful architectures, I'm always looking to collaborate with fellow builders in AdTech and Life Sciences to create real-world value.
Networking
What I can offer
- ›Expertise in taking AI concepts from 0-to-1
- ›Scaling notebook prototypes to production-ready APIs
- ›Adversarial prompt processing and LLM security insights
What I'm looking for
- ›Full-time roles in Generative AI or AI/ML Engineering
- ›Connections with builders in AdTech, Life Sciences, and multimodal systems
Best fit for
Focus
Current interests
Core competencies
Background
Career
Transitioned from Mechanical Engineering to a Python Developer, then Machine Learning Engineer at TCS, followed by a Master's at Northeastern where he served as a Graduate TA for Gen AI. Most recently, he was a Founding AI Engineer at Galexor AI before joining PraxisPro.
Education
Master of Science in Computer Software Engineering, Northeastern University (2023 – 2025); Bachelor of Technology in Mechanical Engineering, Gokaraju Rangaraju Institute of Engineering and Technology (2018 – 2022).
Achievements
- ›Built Galexor MVP in 3 months leading to pre-seed VC funding
- ›Reduced TCS infrastructure costs by 15% and query latency by 30%
- ›Processed 5TB/day of log data at TCS
- ›Published fine-tuned LLaMA models and DDPM framework on Hugging Face
Opinions
- Rigid, stateful architectures are the only way to make LLMs reliable in production.
- Models need professional intuition and logic rather than just more internet data.
- AI is a partner for thoughtful collaboration, not just a tool for prompt-spamming.
Personality
Communication style
Professional, high-energy, and technical; balances academic authority with a builder mentality.
Formality — 7/10
Vocabulary