Experience Projects Skills Education Contact
Available for Opportunities

I'M
SAI SANJEEV

ALPHA TL Manager · NIE Mysore · AI/ML Engineer

CS grad specializing in AI/ML. Deployed 10+ production ML models, optimized data pipelines achieving 25% faster throughput, and led cross-functional teams of 10+ members.

10+
ML Projects
8.7
CGPA
200+
Students Led
Sai Sanjeev
ML Engineer
IoT + AI
Professional Journey

Work Experience

Oct 2024

Machine Learning Intern

Plasmid
  • Developed 3 ML models achieving 25% higher prediction accuracy than baseline
  • Collaborated with 4 cross-functional teams to deliver 20% faster project completion
  • Analyzed 50K+ real-world records, identifying 8 key trends that boosted model precision by 15%
May 2025 –
Present

Student TL Manager

ALPHA, NIE Mysuru
  • Organized 6 workshops and 2 hackathons with 200+ student attendees
  • Mentored 15 junior students, guiding 5 AI-driven projects to completion
  • Launched weekly technical discussions, increasing peer participation by 40%
Oct 2025 –
Present

Teaching Assistant

Interdisciplinary PBL · NIE
  • Delivered 20+ lectures to 80 first-year students on project-based learning
  • Advised 12 student teams, helping 10 secure top grades in final presentations
  • Streamlined documentation templates, reducing report preparation time by 30%
Technical Build Log

Featured Projects

Cybersecurity / Cloud

Secure Cloud File Sharing System

An enterprise-grade web application ensuring complete file privacy using end-to-end hybrid cryptography (AES-256-GCM + RSA-4096-OAEP) and seamless Google Drive integration.

Goal

Solve data privacy concerns inherent in traditional cloud storage by implementing end-to-end encryption. Files must be encrypted before leaving the user's device and only be decipherable by the exact intended recipient.

Architecture

  • Hybrid Cryptography: AES-256-GCM used for fast, secure file encryption. RSA-4096-OAEP encrypts the AES session key to ensure completely secure key exchange.
  • Storage: Encrypted files are hosted on Google Drive, while metadata, user information, and notifications are managed via a MySQL database.
  • Security Layers: Private keys are encrypted using the user's password (via bcrypt and Cryptography Library) before database storage, mitigating potential breaches.

Features

  • End-to-end encrypted file sharing and secure user authentication
  • Read receipt notifications and real-time alerts
  • Multi-user secure communication with encrypted private key storage
  • Direct integration with Google Drive API

Outcome

Successfully developed a production-ready secure file-sharing platform that guarantees data confidentiality, integrity, and privacy while enabling practical cloud collaboration.

AES + RSA Hybrid Crypto
E2E Encryption
Google Drive API
Computer Vision / AI

Car & Pedestrian Detection System

An intelligent, fully offline computer vision system detecting vehicles and pedestrians in real-time, while simultaneously estimating age and gender using YOLOv8m and MiVOLO V2.

Goal

Demonstrate practical implementation of state-of-the-art object detection and demographic analysis locally, ensuring zero cloud dependency and strict data privacy.

Workflow & Architecture

  • Detection Layer: Utilizes Ultralytics YOLOv8m for high-speed, accurate bounding box generation around cars and pedestrians.
  • Demographic Layer: Integrates MiVOLO V2 to estimate age and gender by analyzing facial and body features of detected individuals.
  • Backend: Flask handles image/video streams, passing frames to PyTorch-based inference engines.
  • Resource Management: Features a continuous safety monitoring mechanism (via Psutil) tracking CPU/RAM to dynamically throttle processing, preventing thermal overload during heavy video analysis.

Features

  • Real-time car/pedestrian detection and demographic insights
  • Voice-based detection announcements (Web Speech API)
  • Thermal protection and resource monitoring
  • 100% offline deployment capability

Outcome

Built a comprehensive AI-powered application delivering high accuracy for smart city monitoring, retail analytics, and public safety with highly efficient local hardware utilization.

Real-time inference
100% Offline execution
Thermal protection
Generative AI / NLP

Research Report AI Assistant

A fully offline, privacy-focused AI writing assistant integrating Mistral 7B, Phi-2, and FLAN-T5-base for academic report generation, autocomplete, and grammar correction.

Goal

Support researchers and professionals in creating high-quality academic documents using an intelligent writing environment that processes everything locally, eliminating internet dependence and ensuring absolute data privacy.

Multi-Model Architecture

  • Mistral 7B: Handles generation of complete reports and structured technical content.
  • Phi-2: Powers lightning-fast, real-time autocomplete suggestions.
  • FLAN-T5-base: Executes rigorous grammar checking and spelling correction.
  • Infrastructure: Built on FastAPI with WebSocket communication for zero-latency, real-time text interactions.

Adaptive Learning Engine

The system continuously learns from user corrections, building a personalized writing profile via Pickle serialization. It analyzes vocabulary, sentence structures, and formatting preferences to align perfectly with the user's individual writing style over time.

Outcome

Developed a comprehensive offline AI research assistant capable of handling academic papers, executive summaries, and technical reports while actively adapting to user preferences with guaranteed data security.

3 specialized AI models
Real-time WebSockets
Adaptive learning engine
Cybersecurity / Data Mining

Cybersecurity Attack Analysis System

A comprehensive machine learning platform to analyze network traffic, discover hidden attack patterns via FP-Growth, and classify threats using XGBoost and DBSCAN.

Goal

Go beyond traditional intrusion detection by not only detecting attacks but also uncovering hidden relationships among 85+ network traffic features to provide interpretable, actionable insights for cybersecurity analysts.

Architecture & Workflow

  • Data Mining: Uses FP-Growth association rule mining to discover frequent itemsets and hidden attack patterns from the CICIDS2017 dataset (2.5M+ records).
  • Classification: Employs XGBoost, Random Forest, and Decision Trees to classify attacks (DDoS, Bot, PortScan, etc.). Optimized by selecting the TOP-10 most influential features via XGBoost feature importance.
  • Anomaly Detection: Utilizes DBSCAN clustering with PCA to group similar traffic and flag suspicious, unknown attack behaviors.
  • Deployment: Full-stack Flask/MySQL web application operating completely offline for data privacy, featuring real-time attack predictions.

Outcome

Successfully integrated pattern mining, classification, clustering, and anomaly detection into a unified platform. Achieved high accuracy and faster prediction speeds by reducing 85 features to the 10 most critical ones.

FP-Growth rules
XGBoost + DBSCAN
85 → 10 Features
2.5M+ Records
Industrial IoT

Modbus-Based SCADA System

A low-cost, production-grade SCADA system for real-time industrial monitoring and remote device control — built entirely with off-the-shelf hardware and open protocols.

Goal

Replace expensive proprietary SCADA setups with a cost-effective alternative using commodity IoT hardware. The system needed to be offline-capable, remotely accessible, and maintainable by non-specialist staff.

Architecture

ESP32 acts as the Modbus Slave, collecting DHT11 temperature/humidity readings into mapped Modbus registers. A Raspberry Pi 5 serves as the Modbus Master, polling the ESP32 every 2 seconds via Modbus RTU over serial (UART). The Pi also runs a lightweight local web server (Flask/Python) that exposes a live dashboard with real-time readings and device control buttons.

Implementation

  • Programmed ESP32 in embedded C to sample DHT11 sensor data and populate Modbus register map (holding registers 40001–40010)
  • Configured Raspberry Pi as Modbus RTU master with automatic reconnection and error-handling for dropped packets
  • Built responsive web dashboard (HTML/JS/Chart.js) showing live time-series graphs and toggle controls for connected relays
  • Integrated Cloudflare Tunnel (cloudflared) for zero-trust remote access — no port forwarding, no public IP exposure
  • Added local data logging to SQLite for offline analytics and audit trails

Outcome & Impact

System runs continuously with 99% data accuracy across 72-hour stress tests. Remote access latency under 200ms via Cloudflare Tunnel. Total hardware cost under ₹4,000 vs. ₹50,000+ for commercial equivalents.

Skills Applied

Modbus RTU/TCP, embedded C, Python (pymodbus, Flask), Cloudflare Tunnel, SQLite, IoT system architecture, real-time dashboard development.

99% data accuracy
<200ms remote latency
₹4K vs ₹50K cost
Offline capable
Cybersecurity

STIX Malware Analyser

An automated ML pipeline for malware detection using structured cyber threat intelligence in STIX format — combining unsupervised anomaly detection with supervised classification for layered defence.

Goal

Automate the process of ingesting raw threat feeds, normalizing them into the STIX 2.1 standard, and applying ML models to detect both known and unknown malware with minimal analyst intervention.

Approach

  • Parsed and normalized threat data (IP indicators, file hashes, attack patterns, TTPs) into STIX 2.1 JSON bundles using the stix2 Python library
  • Engineered features from STIX objects: IP geolocation behavior, file signature entropy, lateral movement patterns, C2 communication frequency
  • Layer 1 — Isolation Forest (unsupervised): flags zero-day and novel anomalies where no labelled data exists
  • Layer 2 — SVM with RBF kernel (supervised): classifies flagged samples against a labelled corpus of known malware families
  • Built a report generation module that outputs STIX Course-of-Action objects with remediation suggestions for each detected threat

Results

Hybrid two-layer approach improved overall detection reliability by 30% over a single-model baseline. False positive rate reduced by 18% compared to Isolation Forest alone. End-to-end pipeline processes a 10,000-indicator feed in under 45 seconds.

Skills Applied

STIX 2.1 standard, Isolation Forest, SVM, feature engineering from threat intelligence data, end-to-end ML pipeline automation, Python (scikit-learn, stix2).

+30% detection reliability
-18% false positives
10K indicators / 45s
Predictive Maintenance

NVMe Level-3 Health Analyzer

A predictive maintenance system for NVMe SSDs that uses SMART telemetry data and XGBoost — tuned via Genetic Algorithm — to forecast drive failures before they occur.

Goal

Provide data centre operators and end users with an early warning system for NVMe drive failure, reducing unplanned downtime and data loss. Target: predict failure with at least 90% accuracy with actionable lead time of 24–72 hours.

Workflow

  • Automated extraction of 15+ SMART attributes: NVMe temperature, read/write error rates, power-on hours, unsafe shutdowns, media errors, wear levelling counts, percentage-used endurance
  • Built a custom feature extraction module to handle missing SMART attributes (vendor-specific registers) using median imputation and flag encoding
  • Derived rolling-window features (7-day and 30-day) to capture degradation trends, not just point-in-time values
  • Trained XGBoost classifier — chosen for its native handling of missing data and structured log data
  • Applied Genetic Algorithm (DEAP library, 200+ generations, tournament selection) to optimize XGBoost hyperparameters: max_depth, learning_rate, subsample, colsample_bytree, n_estimators
  • Compared GA-tuned model against Grid Search and Bayesian Optimization baselines — GA achieved best F1 at comparable compute cost

Performance

Processed 10,000 drive records in 137 seconds end-to-end. Achieved 92% failure prediction accuracy with 89% recall on the minority (failure) class. GA tuning added 4.2% accuracy improvement over default XGBoost hyperparameters.

Skills Applied

XGBoost, Genetic Algorithms (DEAP), feature engineering, predictive maintenance, NVMe/SMART telemetry parsing, model evaluation (precision/recall/F1), Python.

92% failure accuracy
89% recall on failures
10K records / 137s
+4.2% from GA tuning
RAG + LLM

AI Document Query Chatbot

A Retrieval-Augmented Generation (RAG) chatbot that answers user questions from uploaded documents using semantic search — not keyword matching — backed by LLaMA 2 and Chroma DB.

Goal

Enable users to query large documents (PDFs, reports, manuals) in natural language and receive accurate, grounded answers — eliminating hallucination by always anchoring responses to retrieved document chunks.

Architecture

  • Documents ingested and split into 500-token overlapping chunks (50-token overlap) to preserve context at boundaries
  • Each chunk converted to a 768-dim dense vector using Sentence Transformers (all-mpnet-base-v2 model)
  • Vectors stored in Chroma DB (local persistent vector store) with document metadata for source attribution
  • At query time: user query embedded with same Sentence Transformer, cosine similarity search retrieves top-3 most relevant chunks
  • Retrieved chunks + user query assembled into a structured prompt and passed to LLaMA 2 (7B, quantized with GGUF) via LangChain's RetrievalQA chain
  • Source citations returned alongside answers, showing which document sections were used

Optimizations

  • Switched from FAISS to Chroma DB — 30% faster retrieval due to persistent indexing
  • Quantized LLaMA 2 to 4-bit GGUF format — reduced memory footprint from 14GB to 4GB, enabling local CPU inference
  • Implemented query caching for repeated questions — reduced average response time from 2.1s to 1.26s (40% improvement)

Metrics

90% answer accuracy on a held-out evaluation set of 200 QA pairs. Retrieval precision@3 of 88%. Runs fully offline on consumer hardware (8GB RAM, no GPU required).

Skills Applied

RAG architecture, vector databases (Chroma DB), LLM integration, LangChain, Sentence Transformers, GGUF quantization, Python.

90% answer accuracy
40% faster responses
2.1s → 1.26s avg
Runs fully offline
FinTech

Online Payment Fraud Detection

A robust fraud detection pipeline for credit card transactions tackling the real-world challenge of extreme class imbalance (only 0.17% fraud cases in 284,807 transactions).

Goal

Build a production-ready fraud classifier that maintains high precision and recall simultaneously — minimising both missed fraud (false negatives, which cost money) and false alarms (false positives, which hurt customer trust).

Process

  • Dataset: Kaggle European credit card dataset — 284,807 transactions, 492 fraud cases (0.17%), 28 PCA-transformed features + Time + Amount
  • Applied SMOTE (Synthetic Minority Oversampling Technique) to oversample fraud class from 492 → 10,000 synthetic samples, preserving realistic feature distributions
  • Feature engineering: transaction velocity (frequency per hour per card), balance shift ratio, time-since-last-transaction, normalized Amount (RobustScaler to handle outliers)
  • Trained and benchmarked three models: Logistic Regression (baseline), XGBoost, Random Forest with 5-fold stratified cross-validation
  • Optimized prediction threshold (default 0.5 → 0.35) using Precision-Recall curve to maximize F1 on imbalanced test set
  • Built a lightweight inference pipeline: feature transformation → model predict_proba → threshold check → alert flag, achieving 25% faster runtime than naive sklearn pipeline

Outcome

Random Forest outperformed both baselines. Final model: 95% precision, 93% recall, F1-score 0.94 on held-out test set. ROC-AUC of 0.98. Pipeline throughput: 50,000 transactions processed per minute.

Skills Applied

SMOTE, imbalanced learning, threshold optimization, Random Forest, XGBoost, feature engineering, precision/recall tradeoff analysis, Python (scikit-learn, imbalanced-learn).

95% precision
93% recall
0.98 ROC-AUC
25% faster pipeline
Smart Agri · Active 🌱

SOIL — Sustainable Organic Intelligence Layer

An end-to-end smart agriculture system providing real-time soil health monitoring and AI-powered crop/fertilizer recommendations for small-scale farmers — a Govt. of Karnataka Grassroot Innovation 2025 finalist.

Goal

Eliminate the dependence on expensive and infrequent laboratory soil testing for small farmers. Provide continuous, affordable, and actionable soil intelligence directly in the field, reducing input costs and improving yield decisions.

System Components

  • Sensor layer: capacitive soil moisture sensor, analog pH probe (SEN0161), NPK sensor (RS485 Modbus output) — all connected to an ESP32 microcontroller
  • ESP32 aggregates multi-sensor readings every 5 minutes and transmits compressed packets over LoRa (SX1278, 433 MHz) to a central gateway up to 2km away
  • Gateway (Raspberry Pi 4) receives LoRa packets, decodes sensor values, and runs local AI inference — no internet dependency
  • AI model: Random Forest classifier trained on a regional crop-soil dataset (5,000 labelled samples across Karnataka crops) — classifies soil health into 3 categories (Healthy / Nutrient-deficient / Degraded) and recommends suitable crops + fertilizer ratios
  • Results displayed on a local e-ink display at the gateway node and a mobile-optimised web dashboard accessible on the farm's local WiFi

Pilot Results

Tested on 3 farms in the Mysore district over 8 weeks. Reduced manual soil testing time by 80% (from 2 hours per sample trip to 20 minutes per week of dashboard review). Crop recommendations matched expert agronomist advice in 85% of cases. Avg. sensor power draw: 12mA — projected battery life of 6 months on a 10,000mAh pack.

Planned Upgrades

  • Weather API integration (OpenWeather) to factor rainfall forecasts into irrigation advice
  • Auto-irrigation control via relay-controlled solenoid valves triggered by soil moisture thresholds
  • SMS alert gateway for farmers without smartphones

Skills Applied

LoRa communication, ESP32 embedded programming, Modbus RS485, edge AI inference, Random Forest, agricultural domain knowledge, low-power IoT design, Raspberry Pi, Python.

80% less manual testing
85% expert match rate
2km LoRa range
🏆 Gov. Finalist 2025
Technical Toolkit

Core Competencies

Languages
Python (advanced)Java CC++SQL
Web & Cloud
FastAPIFlask REST APIsWebSockets Google Drive APIHTML / CSS / JS
ML / Generative AI
PyTorchYOLOv8 Mistral 7B / Phi-2Scikit-learn XGBoostPandas / NumPy
Cybersecurity & Hardware
AES-256 / RSA-4096E2E Encryption Raspberry PiESP32 Modbus / LoRa
Tools & Platforms
Git / GitHubDocker Chroma DBCloudflare
Interests
Offline AI DeploymentDrone Building Competitive Robotics
Academics & Recognition

Education & Certifications

Degree

B.E. Computer Science (AI & ML)

National Institute of Engineering, Mysuru
Aug 2022 – Present
8.7 / 10
Core: Data Structures · Algorithms · ML · Deep Learning · Embedded Systems
Certifications
🏅 Certificate of Merit – Techkriti'24 (top 5% of 500 teams)
🌱 Finalist – Grassroot Innovation 2025, SOIL project (Govt. of Karnataka)
🤖 Generative AI Specialist – NVIDIA DLI Certified (Sep 2024)
📜 Internship Completion Certificate – Plasmid (Oct 2024)
Get in Touch

Let's Connect

Contact Details

+91 80737 08129
Bellary, Karnataka, India — 583101
Languages
English (professional) Kannada (native) Telugu (native) Hindi (basic)