NS
Nishant OS
Initializing Nishant AI
LOADING0%
Open to AI / ML engineering roles & internships

Building AI systemsthat think, reason,and solve real problems.

Nishant ShahAI EngineerMachine Learning EngineerComputer Vision EngineerDeep Learning EngineerFull-Stack AI DeveloperResearcher

Building AI systems that think, reason, and solve real problems. I ship multi-agent LLM frameworks, retrieval platforms, and computer-vision models — from research to production.

0
AI & ML projects
0%
Best model precision
0
Peer-reviewed paper
0
Certifications
Scroll
01About
nishant@os — ~/about
02Capability

A knowledge graph, not a checklist.

Everything traces back to Artificial Intelligence. Hover a node to see its connections; click to see where it ships.

Artificial IntelligenceComputer VisionOpenCVDeep LearningPyTorchTensorFlowscikit-learnLLMsAI AgentsRAGGenerative AILangGraphFastAPIReact / Next.jsTypeScriptDockerKubernetesAWSRaspberry Pi / EdgePostgreSQLGenome AssemblyPython
FoundationsMachine / Deep LearningComputer VisionLLMs & AgentsBackend & WebInfra & EdgeData & Bio

Select any node in the graph to explore how that skill maps to real projects and outcomes.

03Selected Work

Each project is its own product.

Not cards — full systems. Problem, architecture, and measured impact for the work I'm proudest of.

ProductionAgentic Financial OS · Fintech · Jul 2026 – Present

LedgerMind

An agentic financial OS — a virtual accountant, bookkeeper, analyst, and CFO in one.

108
Live backend tests
10/11
Roadmap complete
Deterministic
Numbers
Swappable
LLM providers
PythonFastAPILangGraphPaddleOCRNext.jsTypeScriptSupabase
System Architecturelive flow
Receipt / Gmailupload · OAuthPaddleOCRparseHuman Reviewlow-confidenceCategorizerLLM classifiesFraud DetectionLedger + CFOSupabase · SQL
ProductionESG Document AI · LLM Evaluation · Jun 2026 – Sep 2026

ESGenuine

Evidence-grounded ESG claim extraction — with a deterministic layer that repairs what the LLM gets wrong.

96.1
Precision composite
+24.6
Repair layer gain
0.45 ms
Repair latency
1,730
Claims extracted
PythonFastAPIDoclingLlama 3.3 70BSupabasepgvectorReact
System Architecturelive flow
ESG Report PDFBRSR · IRDocling Parsersections · tablesLLM ExtractionLlama-3.3-70B · NIMDeterministic RepairS1–S5 · 0.45 msSupabasepgvector · provenanceContradictionsNLIIntegrity Reportdashboard
ProductionAI Banking Assistant · Multi-Agent · May 2026 – Jul 2026

NeuroBank

An AI banking assistant that runs real banking operations through conversation.

4
Specialized agents
4
Core workflows
FastAPI
Backend
Secure
Auth
PythonFastAPISQLiteLLMsLangGraphFunction CallingMicroservices
System Architecturelive flow
User Querynatural languageNLU AgentintentPlanner Agentsequences taskReflection AgentverifiesExecution Agentfunction callingBanking APIsFastAPI · SQLite
ProductionLocal-First Knowledge Management · Feb 2026 – May 2026

ClarityStack

Turn long-form AI chat histories into structured, version-controlled knowledge.

3
Microservices
2
Client platforms
Immutable
Ground truth
SRS
Retention
PythonFastAPISQLiteNext.jsTypeScriptJetpack ComposeMarkdown
System Architecturelive flow
AI Chat HistoriesExtractorstructureVersioned KB+ immutable rawSRS ServiceEditor ServiceThreatLensanalyticsWeb + AndroidNext.js · Compose
ResearchMedical Computer Vision · Deep Learning · Sep 2025 – Jan 2026

Diabetic Retinopathy Detection

Grading diabetic retinopathy from retinal fundus images at clinical agreement.

90.42%
Accuracy
0.94
Quadratic Kappa
3
Backbones
Grad-CAM
Explainability
PythonPyTorchTensorFlowEfficientNetConvNeXtSwin-TinyGrad-CAM
System Architecturelive flow
Fundus ImagePreprocesscrop · normalizeBackbonesEfficientNet · ConvNeXt · Swin5-Class GradeGrad-CAMexplainability
ResearchAssistive AI · Edge · IoT · Mar 2024 – Sep 2024

DRISHTI

A Raspberry Pi wearable that guides the visually impaired with real-time multimodal sensing.

98.42%
Detection precision
25ms
Latency
100%
Offline
Multimodal
Sensing
PythonYOLOv8-nanoOpenCVRaspberry Pi 5PyTorchKNNUltrasonic Sensors
System Architecturelive flow
Camera + UltrasonicYOLOv8-nanodetectionFace RecognitionKNN on-deviceImage CaptioningMultimodal FusionpriorityVoice + Haptic
04Research Arc

Six years of curiosity, one trajectory.

From combinatorial biology to reasoning agents — each stop compounding into the next.

05Publications

Peer-reviewed, and reproducible.

Research that shipped with code and results others can verify.

Conference Paper2026

ICTIS 2026 — Int'l Conference on ICT for Intelligent Systems, Bangkok, Thailand · N. Shah, et al.

Abstract

DRISHTI is a Raspberry Pi 5-based wearable edge-AI and IoT system that fuses multimodal sensing — YOLOv8-nano vision, ultrasonic ranging, on-device face recognition, and image captioning — to deliver real-time, fully offline navigation assistance for visually-impaired users. The system reaches 98.42% detection precision at 25 ms latency and delivers guidance through multimodal voice and haptic feedback, eliminating cloud dependence.

·Code
06Recognition

A wall of proof.

First Place — Hackathon

Winner of the hackathon organized by NitroStack and WeKan

Published at ICTIS 2026

DRISHTI presented at the ICTIS 2026 Conference, Bangkok, Thailand

Data Fellowship 2025

Selected for data-driven research and civic-tech innovation

Chancellor's Scholarship

Awarded by Amrita Vishwa Vidyapeetham

5 Professional Certifications

DataCamp, AWS Academy, Udemy, and IIRS/ISRO

07Education & Credentials

The foundation.

A CSE (AI) degree in progress, backed by industry certifications from AWS, DataCamp, ISRO/IIRS and more.

Education
2023 – Present

B.Tech in Computer Science Engineering (AI)

Amrita Vishwa Vidyapeetham, Amritapuri

Specializing in Artificial Intelligence — ML, deep learning, computer vision, and full-stack AI systems.

2020 – 2022

Higher Secondary (+2), Computer Science

Bal Kalyan Vidhya Mandir Secondary School

Certifications
Associate Data Analyst
DataCamp
Feb 2026
Exploring AI: Use Cases & Applications
AWS Academy
Sep 2025
AWS Academy Cloud Foundations
Amazon Web Services
Aug 2025
Machine Learning A-Z: AI, Python & R
Udemy
May 2025
AI/ML for Geodata Analysis
IIRS, ISRO
Sep 2024
09Connect Protocol

Let's build something that reasons.

Open to AI / ML engineering roles and interesting collaborations. Pick a channel, or transmit a message directly.

secure channel · open

Goes straight to my inbox · or use the channels on the left.