class JagrutJoshi:
name = "Jagrut Joshi"
role = "AI Systems Engineer"
university = "DY Patil International University, Pune"
builds = [
"Perception systems — vision + contextual reasoning",
"Safety pipelines — anomaly detection + behavioral AI",
"Health intelligence — emotional context + long-term memory",
]
principles = [
"Systems > Models",
"Context > Detection",
"Decision > Prediction",
"Local AI > Cloud dependency",
]
motto = "Intelligence is not prediction. It is decision-making." |
|
|
🧠 SYSTEMS End-to-end pipelines. |
🔍 CONTEXT Interprets environments. |
⚡ DECISION Actionable intelligence. |
🔐 PRIVACY On-device inference. |
🌍 IMPACT Ships to production. |
flowchart LR
subgraph S["📡 SENSE"]
A[Raw Input]
B[Preprocessing]
end
subgraph U["🧠 UNDERSTAND"]
C[Perception Engine]
D[Context Analysis]
E[Reasoning System]
end
subgraph R["💾 REMEMBER"]
F[Memory Layer]
end
subgraph Act["⚡ ACT"]
G[Decision Engine]
H[Actionable Output]
end
A --> B --> C --> D --> E --> F --> G --> H
The goal is never a model. The goal is a system — one that sees, understands, remembers, and acts.
Three systems. One philosophy — intelligence that perceives, reasons, and acts.
From detection to understanding. Vision that knows what it sees.
▶ Architecture · Intelligence Layer · Deployment
Gap Solved — Detection asks "What is here?" | NeuroVision answers "What is happening — and why does it matter?"
flowchart TD
subgraph In["📥 INPUT"]
A[Camera / Video Feed]
B[Frame Normalization]
end
subgraph Ve["👁️ VISION ENGINE"]
C[Object Detection — OpenCV RT]
D[Spatial Reasoning — Positional Mapping]
end
subgraph IL["🧠 INTELLIGENCE"]
E[Scene Context Modeling]
F[Temporal Pattern Analysis]
end
subgraph Out["📤 OUTPUT"]
G[Annotated Frames]
H[Confidence Scoring]
I[Decision-Ready Summary]
end
A --> B --> C --> D --> E --> F --> G & H & I
| Capability | Implementation | Signal |
|---|---|---|
| Object Detection | OpenCV · Real-time | Live visual pipeline |
| Scene Understanding | Spatial reasoning · Pattern analysis | Environment-aware context |
| Structured Output | Annotated frames · Confidence scoring | Decision-ready data |
| Temporal Continuity | Frame-by-frame state tracking | Understands change over time |
Deployment Targets · Urban Monitoring · Smart Infrastructure · Situational Analysis
Reactive monitoring isn't safety. SafeNet predicts risk before it escalates.
▶ Architecture · Intelligence Layer · Deployment
Gap Solved — Alerts say "Something went wrong." | SafeNet says "Something is about to go wrong — here's when and why."
flowchart TD
subgraph In["📥 SIGNALS"]
A[Visual Feed]
B[Sensor Streams]
end
subgraph De["🔍 DETECTION"]
C[Rule-Based Engine]
D[Visual Heuristics]
end
subgraph An["🧠 ANALYSIS"]
E[Temporal Sequence Modeling]
F[Multi-Factor Context Assessment]
end
subgraph Ri["⚠️ RISK ENGINE"]
G[Risk Scoring — Threshold Intelligence]
H[Proactive Alert + Situation Interpretation]
end
A & B --> C & D --> E --> F --> G --> H
| Capability | Implementation | Signal |
|---|---|---|
| Anomaly Detection | Rule-based + visual heuristics | Context-aware flagging |
| Behavioral Recognition | Temporal sequence analysis | Tracks pattern evolution |
| Risk Scoring | Threshold intelligence engine | Predicts — doesn't react |
| Situation Interpretation | Multi-factor scene assessment | Explains why it's a risk |
Deployment Targets · Public Spaces · Industrial Safety · Critical Infrastructure
Health data without context is noise. NeuroWell builds understanding over time.
▶ Architecture · Intelligence Layer · Deployment
Gap Solved — Trackers say "What happened today?" | NeuroWell answers "What is happening to you over time — and what does it mean?"
flowchart TD
subgraph In["📥 INPUT"]
A[Conversational Input]
B[Behavioral Signals]
C[Health Metrics]
end
subgraph Lo["🔒 LOCAL AI — Zero Cloud"]
D[Emotional Context Detection — NLP + LLM]
E[LM Studio — Local Inference]
end
subgraph Me["💾 MEMORY + PATTERNS"]
F[Long-Term Memory — Cross-Session Extraction]
G[Health Pattern Analyzer — 10+ Alert Rules]
end
subgraph Out["📤 OUTPUT"]
H[Proactive Alerts]
I[AI-Generated Reports — jsPDF]
J[Adaptive Dashboard — PWA · Offline-Ready]
end
A & B & C --> D --> E --> F & G --> H & I & J
| Capability | Implementation | Signal |
|---|---|---|
| Conversational AI | Local LLM via LM Studio | Fully private · Zero cloud |
| Long-term Memory | Cross-session pattern extraction | Knows your full history |
| Emotional Intelligence | Heuristic NLP + LLM classification | Affective context — not just vitals |
| Health Pattern Detection | 10+ intelligent alert rules | Catches behavioral anomalies early |
| Automated Reports | jsPDF · AI-generated narrative | Synthesizes patterns into insight |
| PWA + Offline | Service Worker · Installable | Production-ready edge deployment |
Deployment Targets · Personal Health Awareness · Preventive Care · Privacy-First Health AI
| 👁️ | Visual Intelligence | Scene-aware perception pipeline | Vision that interprets, not just detects |
| 🛡️ | Safety Systems | Proactive anomaly + behavioral prediction | Catches risk before it escalates |
| 🧬 | Health AI | Conversational AI with long-term memory | Builds longitudinal understanding, not logs |
| 🔒 | Privacy-First AI | Fully local LLM deployment | Production-grade AI · Zero cloud exposure |
| 🏗️ | Workflow Automation | Ticket management · Document generation | Automated decision-support pipelines |
|
|
Most AI projects optimize for benchmark performance. I optimize for real-world decision quality.
A model that scores 94% in evaluation and fails under production conditions is not a solution — it is a proof of concept.
I build systems that reason through ambiguity, handle noise, and hold up in the messiness of the real world.
|
🔬 Context-Aware AI Systems that model environments — not just classify inputs. 👁️ Vision Reasoning Scene-level comprehension. Beyond bounding boxes. |
🧬 Human-Centric AI Longitudinally intelligent, emotionally aware systems. 🔒 Local LLM Deployment Production AI that runs entirely on-device. |
🛡️ Intelligent Safety Systems Proactive risk modeling. Not reactive alerting. ⚡ Edge AI Deployable intelligence — zero latency, zero cloud dependency. |
👁️ NeuroVision AI Scaling to multi-context |
🛡️ SafeNet AI Edge hardware |
🎯 Internship Search AI Systems & ML |
📊 Extended GitHub Insights
|
|
|
flowchart LR
subgraph Done["✅ SHIPPED"]
A[Full-Stack AI Systems]
B[Local LLM — Privacy-First Architecture]
C[Safety + Perception Pipelines]
end
subgraph Now["🔄 IN PROGRESS"]
D[NeuroVision — Multi-Context Scaling]
E[SafeNet — Edge Hardware Integration]
F[AI Systems Internship]
end
subgraph Next["⬜ NEXT"]
G[Unified Vision + NLP Pipeline]
H[Research Publication]
I[Cloud ML Certification]
end
Done --> Now --> Next
"I don't build models. I build systems that see, understand, and act." |