Sanmukh (Sam) Karri
Machine Learning & AI Engineer · San Francisco, CA
I turn messy data into AI that ships — production ML pipelines, computer vision, and GenAI/RAG systems.
I'm an ML/AI engineer and data scientist with 6+ years taking models out of notebooks and into production — recommendation engines, healthcare computer-vision pipelines, and RAG chatbots that real users depend on. Lately I'm deep in LLM agents and generative AI. Outside work you'll find me in the gym, which is exactly why half my side projects are AI coaches.
Open to ML / AI Engineer roles · San Francisco or remote
What I work with
AI / ML
- Machine Learning
- Deep Learning
- Computer Vision
- Generative AI
- LLM Agents
- RAG
- CNN / RNN / GANs
- NLP
- MLOps
Data & Big Data
- Python
- R
- SQL
- PySpark
- Spark / Hadoop
- Databricks
- Snowflake
- Neo4j
- pandas
- TensorFlow
- Scikit-Learn
Build & Cloud
- AWS · Azure · GCP
- Node.js
- React.js
- Flutter
- OpenCV
- Docker
- Git
- ETL
Selected projects
Enterprise Memory Agent
FeaturedA multi-agent enterprise research assistant with persistent memory and autonomous cross-app execution, built on Composio.
RepRight
FeaturedAn AI personal trainer that watches your form in real time — corrects technique and highlights which muscles you’re activating live on screen.
Vitis-AI Porting Advisor
FeaturedScans a model for DPU compatibility and generates two AI-powered refactoring proposals to get it hardware-ready.
An AI trainer that reads posture, fatigue, and injury risk from your voice alone — no camera, no wearables. Powered by Grok / the xAI API.
A backtester that kills bad trading strategies before they cost you money — real data, real costs, out-of-sample splits that catch overfitting.
A distributed AI creative-campaign generator that coordinates a swarm of agents across Akamai + Magnific.
What I'm building lately
Agent Breach Replay is privacy-first: by default, it captures metadata, hashes, labels, and policy decisions, not raw personal data.
SiteVelocity is an Al-powered development intelligence platform that helps real estate developers find the sites most likely to move from opportunity to construction.
Paper-to-Results Graph turns research papers into executable evidence
auto-updates hourly from github.com/samshanmukh
Experience & education
Data Scientist · CLINICOM
Feb 2024 — May 2026Built and deployed ML pipelines for healthcare chatbots — RAG with GenAI foundation models + vector DBs for accurate retrieval, plus computer-vision pipelines (segmentation, detection, anomaly detection) on CNN/transfer-learning architectures.
Data Science Intern · QQ Tech, Inc.
Jun 2023 — Dec 2023Shipped ML/DL models end-to-end, optimized graph algorithms in embedding spaces for pattern recognition, and built NLP analysis of legal documents on Spark / Databricks / Snowflake.
Senior Software Development Engineer · Focus-N-Fly
Jan 2020 — May 2023Designed a personalized recommendation engine (clustering + collaborative filtering) and customer-segmentation models on a Hadoop/Spark stack, from prototype to production.
Software Engineer · Navtech
Dec 2018 — Dec 2019Built and maintained ETL pipelines in Python/SQL over distributed Hadoop/Spark, with data-quality checks and ML support for predictive analytics.
Software Developer & Data Analyst · Prahem Technologies
May 2017 — Nov 2018Built scalable data-processing pipelines and interactive Tableau / Matplotlib dashboards; implemented early predictive models (regression, decision trees).
Education
- University of the PacificM.S. in Data Science · 2022 — 2024
- Jawaharlal Nehru Technological UniversityB.Tech · 2015 — 2019
Languages · English · Hindi · Spanish
Certifications
- ›Introduction to Generative AI — Google Cloud
- ›Attention Mechanism — Google Cloud
- ›Encoder-Decoder Architecture — Google Cloud
- ›Machine Learning with Python — Coursera
- ›Python Essentials for MLOps — Coursera
What people say
A fast learner who isn't afraid to step outside his comfort zone — he regularly raised helpful points and asked the right “why” questions. He'll be an asset to any organization looking for a bright data scientist.
An incredible asset to the dev team. I can always count on him to roll up his sleeves — whether it's debugging or development, he's accountable to the project and his team. A joy to work with.
Dedicated and passionate, always up for challenges and quick to learn. We worked together on a health & fitness project providing automated coach training.
Recommendations from LinkedIn · more on my profile ↗
Away from the keyboard
I lift. A lot. The gym is where I problem-solve away from a screen — and it's the reason half my side projects are AI coaches. Form checks, progressive overload, and the occasional ego lift.
- Favorite lift
- Deadlift
- Training
- 5×/week
- Built for it
- RepRight + VoiceCoach
// yes, two of my projects are literally AI gym coaches.
AI personal trainer — it watches your form and coaches you in Grok's voice
Turn on your camera + mic, stand side-on, and squat — it counts your reps, checks depth, and coaches you out loud in Grok's real voice. Talk back anytime.
form check
—
- › Sees you — pose estimation counts reps + depth (my RepRight)
- › Talks with you — real-time Grok voice (grok-voice-think-fast-1.0), my VoiceCoach
🎧 Use headphones. Camera runs on-device; mic audio streams to xAI to power the live voice.
Let's build something
Hiring, collaborating, or just want to talk shop about agents and computer vision? My inbox is open — I usually reply within a day.