Hello, My Name is
Teja krishna
AI/ML & FULL STACK DEVELOPER | CERTIFIED ETHICAL HACKERAbout Me
I'm Teja Krishna, a B.Tech CSE graduate specializing in AI & Machine Learning from SRM University-AP and a Certified Ethical Hacker (CEH). I enjoy building web applications, exploring cybersecurity, and working with AI/ML — from writing clean full-stack code to training models and securing systems. I'm currently looking for opportunities where I can apply what I've learned and keep growing as an engineer.
Education
Education
Intermediate
Fiitjee Jr College
Jun 2019 - July 2021
Computer Science and Engineering — AI & ML
SRM AP UNIVERSITY - CSE (AI & ML)
Aug 2021 - Sep 2025
Projects
Pro-AI
Full-Stack Multi-Model AI Application- Overview & Action: Developed and deployed a responsive web application, Pro-AI, using React.js, Node.js, and Express.js, showcasing full-stack development expertise. Integrated advanced AI technologies, including ChatGPT, DALL-E, and Gemini AI, enabling dynamic content generation and multi-modal interactivity.
- Result & Impact: Implemented AI-driven functionalities that improved application interactivity, resulting in a 25% increase in user engagement.
Tech Stack:
React.js
Node.js
Express.js
ChatGPT / OpenAI
DALL-E
Gemini AI
REST APIs
Anomaly Detection
May 2025- Implemented a robust data preprocessing pipeline for network traffic analysis using the CIC IDS 2017 dataset, addressing data quality issues and reducing feature redundancy.
- Applied machine learning expertise to implement an anomaly detection system, combining K-Means clustering (Elbow Method) and Random Forest classification (99.96% accuracy) for network security.
- This approach significantly enhanced anomaly detection capabilities, minimizing false positives and demonstrating expertise in addressing real-time cybersecurity challenges.
Tech Stack:
Python
Machine Learning
BB Workspace (BBOS)
Autonomous Multi-Agent AI & Security Decision Platform- Multi-Agent Neural Routing & MCP: Engineered a 34-agent decentralized swarm with dynamic task-affinity routing between Qwen 2.5 Coder (step decomposition & JSON schema structuring) and DeepSeek Coder (AST semantic code analysis & triage) via local Ollama inference and dynamic Model Context Protocol (MCP) function calling.
- Dialectic Consensus & Hallucination Mitigation: Implemented a 3-tier "Debate & Judge" consensus engine where an independent arbiter synthesizes contradictory model outputs. Findings with semantic agreement ≥ 0.70 proceed autonomously, while high-entropy disagreements trigger Human-in-the-Loop gates — reducing false-positive hallucinations by over 60%.
- Dense Vector RAG & Deduplication: Integrated persistent ChromaDB (384-d
all-MiniLM-L6-v2and 768-dnomic-embed-text) with Redis working memory, computing cosine similarity (≥ 0.88) for latent semantic deduplication across thousands of endpoints and dynamic top-k in-context prompt injection. - Graph Reasoning & Dijkstra Optimization: Modeled attack surfaces as Directed Graphs (NetworkX). Mitigated LLM context-window degradation by pairing local LLM semantic transition hypotheses with classical Dijkstra cost optimization (inverting transition probabilities into non-negative additive weights $1/P$) to isolate maximum joint-probability exploit chains.
- Probabilistic Modeling & Continuous Adaptation: Formulated finding likelihood through dynamic Bayesian belief updating across multi-tool evidence, continuous exponential temporal decay ($C(t) = C_0 e^{-\lambda \Delta t}$), and Contextual Multi-Armed Bandits for prompt template optimization with asymmetric agent credit assignment.
- MLOps, Provenance & Zero Cloud Leakage: Delivered 100% on-premise local inference with zero sensitive telemetry leakage, cryptographically auditable SHA-256 prompt lineage, and automated regression benchmark suites achieving ≥ 95% accuracy and ≤ 1.5% false validation rate.
Tech Stack:
Python 3.12
Multi-Agent Swarm (34 Agents)
Ollama (Qwen & DeepSeek)
ChromaDB (Vector RAG)
NetworkX (Dijkstra)
Bayesian & Temporal Modeling
MCP Function Calling
Redis Streams
Pydantic
PyRIT / Garak Red-Teaming
Docker
Skills & Certificates
Languages
Python, Java, C/C++
Web Technologies
HTML5, CSS3, JavaScript, Bootstrap, jQuery
Frameworks & Libraries
React.js, Node.js, Express.js, Scikit-Learn
Cybersecurity & Tools
Penetration Testing, Vulnerability Assessment, Network Security, Web App Security, Threat Analysis, Malware Analysis, Cloud & IoT Security
Google Data Analytics
ViewCertified Ethical Hacker (CEH)
ViewProject in Mind
Let's build your project and give it a life.