📅 Auto-updated daily via GitHub Actions (lowlighter/metrics)
I don't just write ML code — I design the architecture around the problem: where the bottleneck actually is, what breaks at scale, and what the cheapest correct solution looks like. Whether it's a RAG pipeline that needs to stay cheap in production, a booking backend that can't afford race conditions, or a real-time CV system running on a webcam — my job is to find the right-sized solution, not the fanciest one.
- 🏗️ Systems thinker — I design for failure modes (race conditions, cost blowups, latency) before I write the first line of model code.
- 🎯 Full ML lifecycle — data → embeddings/features → training/orchestration → evaluation → deployment, end to end.
- 💸 Cost-aware by default — I default to free-tier, open-weight, and efficient architectures unless there's a real reason not to.
- 🏆 2× Hackathon Prizewinner — shipped full products from zero in 30-hour sprints, twice, under real judging pressure.
- 🧑🏫 Taught what I know — ran a hands-on LLM-tooling workshop for 20+ junior developers, not just built solo.
Cost-Efficient RAG System — a retrieval pipeline (ChromaDB + LangChain + HuggingFace embeddings + Groq LLaMA-3) designed explicitly around minimizing per-query cost without sacrificing retrieval quality — not just "a RAG demo," but a cost model for one. 🔗 github.com/Naveenp7/Cost-Efficient-RAG
LLM-as-Judge Evaluation Pipeline — an automated evaluation system with explicit bias mitigation for position bias, verbosity bias, and sycophancy — the kind of infrastructure that makes LLM outputs trustworthy enough to ship. 🔗 github.com/Naveenp7/llm-judge
Movie Seat Manager — a high-concurrency booking backend (.NET 8, PostgreSQL, Redis) solving double-booking with distributed locking, idempotency keys, and ACID-compliant transactions. Systems design that generalizes directly to e-commerce and SaaS at scale. 🔗 github.com/Naveenp7/Movie-Seat-Manager
Enterprise RAG Knowledge Assistant — end-to-end document Q&A pipeline (LangChain, ChromaDB, Sentence Transformers, FastAPI) covering chunking, embedding, retrieval, and prompt orchestration as a deployable service, not a notebook.
Crowd Detection & Density Estimation — real-time YOLO-based video analytics for public-safety and smart-city use cases, with configurable alerting — built for production constraints, not just accuracy on a benchmark. 🔗 github.com/Naveenp7/Crowd-Detection
AI Resume Screener — BERT-embedding-based semantic matching engine ranking candidates against job descriptions, with a full NLP pipeline for PDF parsing, NER, and skill extraction.
Languages: Python · JavaScript/TypeScript · SQL · Java · Dart
ML & AI: PyTorch · TensorFlow · Keras · Scikit-learn · XGBoost · HuggingFace Transformers · LangChain · NLTK · spaCy
LLM Engineering: RAG Pipelines · Prompt Engineering · LLM-as-Judge · Agentic Workflows · GPT · BERT · Claude · Groq
Backend & Systems: FastAPI · Flask · Node.js · Express.js · REST API Design · Distributed Locking · Auth Workflows
Data: PostgreSQL · Firebase Firestore · Redis · SQLite · Pandas · NumPy · SHAP
Cloud & DevOps: Docker · Kubernetes (fundamentals) · Vercel · GCP · Model Deployment
Frontend: React.js · Next.js · Tailwind CSS · Flutter/Dart
- 🥈 2nd Prize — MATRIX Hackathon (30-hour), CSE Dept., MES College of Engineering — shipped a complete full-stack product from zero, ahead of 20+ competing teams.
- 🥉 3rd Prize — KOTECH Tech Hackathon (July 2025), Qismat Foundation × Kottakkal Municipality.
- 🎓 Workshop Facilitator (ADTEC Program) — led a hands-on AI coding workshop for 20+ junior developers on LLM tooling, prompt engineering, and AI-assisted dev workflows.
- ☁️ Google Cloud Skill Badges — Docker, Kubernetes, IAM, Cloud Storage, Monitoring; Compute Engine fundamentals.
Open to AI/ML Engineer, AI Engineer, and software architecture-adjacent roles — anywhere in India or remote. If you're solving a hard retrieval, evaluation, or systems-design problem, I'd love to hear about it.
📧 naveensanthosh830@gmail.com · 🔗 LinkedIn · 🌐 Portfolio
English (Professional) · Malayalam (Native) · Hindi & Tamil (Conversational)