Nitigya Kargeti
Applied ML Engineer | Building Production AI Systems
I build AI systems that go from research prototype to production. Currently automating data pipelines with LLMs at Dataplatr. Previously: engineered LLM-assisted educational robots deployed to 20 families (ACM CHI 2025). MS in Data Science from UW-Madison, focused on Human-AI Interaction and Brain-Computer Interfaces. My work bridges research and engineering—taking cutting-edge ML and making it work at scale, in production, for real users.
Experience
Professional journey spanning research institutions, startups, and government organizations, focusing on scalable systems, machine learning, and innovative technology solutions.
Data Scientist / Data Engineer
Dataplatr Consulting • Remote (USA)
Engineers spent 8-10h manually writing boilerplate SQL per table for Medallion pipelines → built LLM-driven code-gen tool (Databricks DLT + GPT-4) that auto-drafts L0→L3 queries with metadata enrichment → reduced effort to 3-4h review (60% time saved) across 3 Oracle/BQ schemas (1M rows, 400 MB)
Graduate Research Engineer
People & Robots Lab (NSF-funded), University of Wisconsin–Madison • Madison, WI
2nd Author, ACM CHI 2025: Wizard-of-Oz robot demo wasn't scalable for real homes → engineered full software & AI stack (GPT-4-o VLM + RAG grounding + Python/JS SDK) turning demo into deployable platform → deployed to 20-family in-home study preventing hallucinations via JSON page-context grounding
Software Engineer
Spenza Inc. (B2B SaaS, ~8-person) • Remote / Bengaluru, IN
Platform only supported hard-coded INR currency, limiting global expansion → built dynamic multi-currency module (NestJS + MongoDB + Stripe API) with auto-FX rates and legacy-row migration → shipped 3 PRs with zero regressions, enabling currency toggle for 500+ clients
Projects
A collection of impactful projects spanning web development, machine learning, systems programming, and AI applications, each solving real-world problems with innovative approaches.
Chicago Crimes Forecasting & Hotspot Analysis
Personal Project • 2024
Built GCP Medallion lakehouse processing 10-yr Chicago crime data (7.2M records); achieved 74.8% storage reduction (CSV 446.7 MB → Parquet 112.8 MB), 10× query speedup, H3 spatial clustering identified 85k crime hotspots with 99% geocoding success across resolution levels r7-r9. Architecture: BigQuery + Dataform ETL (Bronze→Silver→Gold medallion), H3 hex-binning spatial aggregation, DBSCAN clustering, Streamlit dashboard.
Product Semantic Search Copilot
Personal Project • 2024
Built hybrid semantic search indexing 120k products + 3M Amazon reviews; improved nDCG@10 by 18% (0.74→0.87), MRR@10 by 21% (0.68→0.82) on benchmarks; achieved 500 queries/min @ 2GB RAM (1-2 CPU), 300ms hybrid-only latency, 800ms with re-rank (10-20% faster on local Docker vs HF Spaces). Architecture: BM25 + BGE-small embedding index + cross-encoder re-ranker; FastAPI backend + Streamlit UI; Docker containerization with /healthz endpoint, structured JSON logs, basic rate-limiting; deployed on HF Spaces + Cloud Run.
Portfolio AI Mode — Resume Q&A + Hybrid Search (Voice & Text)
Personal Project • 2024–2025
Deployed voice + text portfolio Q&A assistant showcasing personal projects/experience; achieved 150-200ms NLP-only latency, 0.8-1.2s LLM-enhanced response time; handled ≈ 50 concurrent sessions with <1% error-rate; 92% intent-detection accuracy, 88% content-relevance on bench-tested queries; attracted ≈ 500 beta visits. Architecture: Groq LLaMA-3.1-8B / Qwen-2.5-8B FP8 via open APIs; hybrid query-processor (keyword retrieval + LLM-reasoning); FastAPI Docker backend on HF Spaces, Next.js frontend on Vercel.
Education
M.S., Data Science (Professional, non-thesis)
University of Wisconsin–Madison
GPA: 3.55/4.0
B.Tech., Computer Science (Minor: Computational Intelligence & Deep Learning)
Manipal University Jaipur
GPA: 10.0/10.0
Skills & Technologies
Programming Languages
Backend & APIs
Databases
Big Data & Analytics
DevOps & Cloud
Publications
Research contributions in brain-computer interfaces, machine learning, and educational robotics, published in peer-reviewed journals and conferences.
Blog
Insights and experiences from building intelligent systems, modern web applications, and production machine learning solutions.
Building This Website: From Figma to Production
A comprehensive journey through designing, developing, and deploying this modern portfolio website with AI-powered features. From Figma wireframes to production deployment, covering component architecture, AI integration, performance optimization, and lessons learned.
Get In Touch
I'm always interested in hearing about new opportunities, especially those involving challenging technical problems.
📍 Currently in Palo Alto, CA