ML systems, retrieval, backend engineering

Meet Gandhi builds production-minded AI projects with measurable systems depth.

Computer Science graduate student at RIT with experience across PyTorch, FAISS, FastAPI, AWS, Docker, data pipelines, and research-grade evaluation.

View projects GitHub LinkedIn

Featured project system

Portfolio projects with live demos, metrics, and case-study depth.

Live GitHub project

SIMBA

Scalable Similarity Search Benchmarking Architecture

A chess decision-space retrieval system using FAISS-style nearest-neighbor search, evaluation-aware embeddings, and systems benchmarks for large position corpora.

ml retrieval backend data

10M position target Sub-second retrieval goal Ranking evaluation

Repository Demo

Next project

Pharmacy Document Intelligence

Public-safe rebuild in progress

A synthetic-document ML platform for semantic search, hybrid retrieval, reranking, explainable evidence views, and production-style FastAPI workflows.

ml retrieval backend

Transformer retrieval Hybrid ranking Docker + CI

Repository Demo

Planned upgrade

Poker Behavior Modeling

Predictive strategy analytics

A modeling and visualization system for poker action prediction, player behavior features, calibration, simulation, and interactive decision analysis.

ml data backend

Behavior features Action prediction Interactive dashboard

Repository Demo

Research signal

Augmented KGE Research

Knowledge graph embedding evaluation

Research pipeline for knowledge graph embeddings, negative sampling, link prediction, and IR-style relevance metrics including MRR, Hits@K, nDCG, and MAP.

ml retrieval data

KGE models IR metrics PyTorch pipelines

Repository Demo

Computer science coverage

Each project is chosen to compound across ML and mainstream engineering roles.

Machine learning

Modeling, retrieval, reranking, evaluation, calibration, and model-card discipline.

Software engineering

Typed APIs, CI, tests, clean architecture, deployable demos, and readable case studies.

Systems and data

Vector search, indexing, pipelines, caching, observability, and dataset safety.

Product signal

Interactive demos that make the project understandable to recruiters and engineers.

Live engineering signals

Small moving parts that make the site feel like a working ML lab.

01 0

Portfolio projects tracked in the manifest.

02 0

Active role tracks: SWE, ML/AI, backend/data.

03 0

Hour maintenance window for project improvements.

Plan Build Verify Deploy Maintain

Rollout cadence

Portfolio first, then a new approved ML project every three days.

Now Launch portfolio hub
Deploy domain, add current GitHub projects, and establish the manifest system.
Next Pharmacy document intelligence
Public-safe synthetic document retrieval with semantic search, reranking, and API demo.
Then Poker behavior modeling
Upgrade or rebuild poker analytics into a predictive ML and strategy simulation project.
Ongoing Meaningful maintenance
Randomized daily commits only when they improve tests, docs, benchmarks, or demos.