Dhadi Sai Praneeth Reddy on Tal Headline: Co-Founder, CEO & AI Researcher @Atlas AI Labs | Founding Member @Botella AI Labs | AI Engineer @Techmatic Systems | Building Sovereign Agentic AI from India | LLMs • Agentic AI • AI Architecture About: Hi, I'm Dhadi Sai Praneeth Reddy, a 4th-year CS undergraduate at Vasavi College of Engineering, Hyderabad, building at the intersection of Agentic AI, Multi-Agent Systems, Cognitive Security, and Applied LLM Research. I don't just study AI. I architect it, train it, and deploy it at government scale. WHAT I'VE SHIPPED 1. Eagle Force Telangana (India's First AI-Integrated Law Enforcement Platform): Solo-architected and deployed eagleforce.tspolice.gov.in. Built three production AI systems (Sahay.AI, Shield.AI, MITRA-TG) and EagleAI, a proprietary 3.1B parameter transformer LLM built from scratch with custom architecture, pretraining pipeline, and tokenizer. Launched by DGP of Telangana. Recognized by the Chief Minister of Telangana. 2. QwiXGenie, 600,000+ Learners: Architected an LLM-powered personalized coding tutor for Tier-2 and Tier-3 cities across India. Featured as Chapter 26, Impact India AI Summit 2026 AI in Education Casebook. 3. T-RACE ONE: Architected India's first sovereign OSINT-based identity intelligence platform. Discovers, correlates, and analyzes digital presence across 2,003+ platforms (1,168+ India-specific) in under 60 seconds. Built with encrypted data processing, DPDP Act 2023 compliance, and on-device intelligence workflows for law enforcement, cybercrime units, and threat intelligence teams. WHAT I'M RESEARCHING 1. AOLP (Agent Ontogeny & Lineage Physics): Original framework reframing AGI development as irreversible physical phase transitions. Introduces Developmental Gating, Irreversibility Index (DII), and Law of Compressive Necessity, a formal alternative to scaling-law paradigms for structurally stable intelligence. 2. KiranaChain: Multi-agent negotiation benchmark on India's 12 to 14 million store informal retail economy. Supports offline RL, MARL, and GNN research. HuggingFace, DOI: 10.57967/hf/9015. 3. NeuroGenesis: Self-optimizing compiler combining Transformer IR encoding with reinforcement learning. Achieves 0.28 execution cost reduction vs. 0.00 LLVM baseline. Outperforms MLGO and AutoPhase. 4. GCT-100K: 100,000-sample cognitive threat benchmark introducing the Cognitive Manipulation Score (CMS). Achieves 84.8% classification accuracy on social engineering and OTP fraud detection. ATLAS AI LABS An independent student research collective, 3 engineers and 1 faculty mentor, building open and reproducible AI research across AGI theory, multi-agent systems, SLMs, and cognitive security. Currently open to internships and full-time roles where the work is serious and the problems are hard. Location: Hyderabad AI stack: - claude code - mcp - Milvus - Pinecone - Weights & Biases Weave - Weaviate Work experience: - AI Engineer at Techmatic Systems India Pvt Ltd, Jul 2026 - Present · 4 mos, Hyderabad, Telangana, India, Joined as AI Engineer and shipped a production-ready agentic platform, a portable offline AI system, a 1M-row enterprise dataset, and a full company website + SEO overhaul — all within the first three weeks. - Built FlowOS, a full-stack Agentic Workflow Builder (visual designer, multi-agent orchestration, connector framework, execution engine). Validated across 400 automated/manual test cases at 100% pass rate; packaged as a Windows desktop app (Electron.js) with a zero-touch auto-installer for Python/uv dependencies. - Built AirDrive, a privacy-first portable AI platform running a local LLM, RAG pipeline, and web search entirely from an encrypted external drive — completed end-to-end with zero host-system data footprint. - Engineered a 1,000,000-record, 224-attribute synthetic enterprise dataset via a custom data-generation pipeline built from scratch; published to Hugging Face with full documentation for AI training and IP licensing. - Rebuilt and deployed the company's production website, then delivered technical SEO across 25 pages: structured data (JSON-LD, FAQ/Product schema), GA4, Google Search Console, and Microsoft Clarity integration — shipped with zero console errors. - Evaluated cloud (Claude API) vs. local-LLM architecture for the core platform and drove adoption of a tool-calling-first design, reducing AI inference dependency and operational cost. - Selected to the founding core team of Botella AI Labs; own a research charter spanning Cybersecurity AI, Agentic AI Safety, and proprietary LLM/SLM architecture — currently authoring the Methodology and Related Work sections of a Cybersecurity AI research paper. - Co-Founder & AI Researcher at Atlas AI Labs, Jan 2026 - Present · 10 mos, Hyderabad, Telangana, India, Co-Founder & Lead Researcher at Atlas AI Labs - student-led AI research lab at Vasavi College of Engineering, Hyderabad. Building agentic systems, sovereign AI tools, and open research infrastructure. Flagship Product T-RACE ONE - India's first sovereign OSINT identity intelligence platform. Correlates digital presence across 2,003+ platforms (1,168+ India-specific) in under 60 seconds. DPDP Act 2023 compliant, on-device, built for law enforcement and threat intelligence teams. Research AOLP - Theoretical framework reframing AI development as irreversible physical phase transitions. Introduces Developmental Gating, Irreversibility Index (DII), and Law of Compressive Necessity toward ontogeny-driven AGI. KiranaChain - Multi-agent negotiation benchmark on India's 12–14M Kirana store economy. Models informal credit, supply-chain dynamics, and multilingual commerce. HuggingFace release · DOI: 10.57967/hf/9015. NeuroGenesis - Self-optimizing compiler using Transformer IR encoding + RL. Autonomously discovers optimization rules. 0.28 execution cost reduction vs. 0.00 LLVM baseline. Outperforms MLGO and AutoPhase. GCT-100K - 100,000-sample cyber manipulation benchmark (phishing, OTP fraud, social engineering) with novel Cognitive Manipulation Score (CMS) metric. Apache 2.0. Stack: Agentic AI · Transformers · RL · OSINT · GNNs · LLM Research · Sovereign AI - Full Stack AI Developer Intern at Telangana Police, Apr 2026 - Jul 2026 · 4 mos, Hyderabad, Telangana, India, Building production-grade infrastructure for Telangana's state emergency response ecosystem under the Telangana Integrated Command & Control Centre (TGICCC). Key Contributions: TGICCC Operational Portal & Dial 112 Emergency Platform - Architected and developed the primary operational portal and Dial 112 emergency web application, with responsive interfaces integrated into high-throughput backend services handling real-time public safety data. AI/ML Routing Protocols — Designed machine learning-driven networking protocols to optimize data dispatch pipelines between Public Safety Answering Points (PSAPs) and field officers, with predictive modules dynamically prioritizing high-urgency traffic during critical incidents. Latency Optimization — Reduced emergency response contact time through low-latency backend logic, asynchronous queue management, and optimized API endpoints across multi-agency coordination workflows. Infrastructure — Production-grade deployment utilizing real-time WebSockets, telemetry logging, and secure state management for live emergency communication health monitoring. Tech Stack: Python · Node.js · React.js · WebSockets · RESTful APIs · Tailwind CSS · Predictive Routing · Low-Latency Architecture · Emergency Response Systems - AI Engineer Intern at ELITE ACTION GROUP FOR DRUG LAW ENFORCEMENT, Jun 2025 - Mar 2026 · 10 mos, Hyderabad, Telangana, India, Architected and deployed eagleforce.tspolice.gov.in - India's first AI-integrated law enforcement platform as a Solo developer, from first line of code to live production on a eagleforce.tspolice.gov.in domain. Delivered a 26-module full-stack ecosystem under the guidance of DIG Shri Abhishek Mohanty IPS and SP Shri KCS Raghu Vir, formally launched by DGP Shri B. Shivadhar Reddy IPS on March 31, 2026. Three production AI systems engineered end-to-end: 1. SAHAY.AI — 24×7 multilingual conversational AI (English, Telugu, Hindi, Urdu) providing anonymous real-time support for substance abuse intervention. Built with structured guidance workflows for user safety at scale. 2. SHIELD.AI — Predictive risk assessment engine using behavioral pattern analysis to generate personalized prevention strategies and enable early-stage intervention through adaptive response mechanisms. 3. MITRA-TG — Specialized adolescent screening tool for parents and educators, built with confidentiality-first design and multilingual accessibility for early detection of substance use risk. - Custom LLM: Sole architect and trainer of EagleAI — a proprietary 3.1B parameter transformer model built from scratch (custom architecture, pretraining pipeline, and tokenizer), trained exclusively on Indian drug law enforcement data, TGANB policy documents, and Eagle Force operational knowledge. - Tech Stack: React 18, TypeScript, Vite, Tailwind CSS, Node.js, Express.js, MongoDB Atlas, i18next, React Query, Radix UI, PDFKit, Nodemailer Recognition: 1. Felicitated by Hon'ble Chief Minister Shri Anumula Revanth Reddy — "Anti-Drug Awareness Soldier" award, Jun 26, 2025 2. Recognized by DGP of Telangana Shri B. Shivadhar Reddy IPS at official platform launch, Mar 31, 2026 - Co-Founder at QwikZen Group India, Nov 2023 - Jan 2026 · 2 yrs 3 mos, Hyderabad, Telangana, India, Co-founded and led QwikZen Group India as CTO, building a suite of production-grade Generative AI platforms serving learners and developers across India. Scaled QwiXGenie to 600,000+ enrolled learners within months of launch, with featured recognition as Chapter 26 in the Impact India AI Summit 2026 AI in Education Casebook. Three flagship AI platforms built and shipped: QwiXGenie — AI-powered personalised coding and technical tutoring platform targeting Tier-2 and Tier-3 cities in India. Integrated LLM-based tutors fine-tuned on curated programming corpora, with adaptive engines that decompose learner queries into conceptual gaps, generate multi-path solutions, and surface debugging strategies over final answers. Partnered with training institutes and schools for institutional reach. Governance built with minimal-data-collection policies, encryption at rest and in transit, parental dashboards for minors, and content moderation layers. QwiX CV — AI-driven resume optimization platform with real-time ATS scoring, job description matching, and automated job visibility enhancement for active job seekers. QwiX Shield — AI platform for drug recovery and prevention support, providing anonymous, accessible assistance to individuals affected by substance abuse. Impact: QwiXGenie case study published in the Impact India AI Summit 2026 AI in Education Casebook (Chapter 26) — recognized alongside national-scale EdTech initiatives for consumer-grade AI tutoring with safety and accountability as core design principles. Education: - Bachelor of Technology - BTech, Computer Science Engineering, Vasavi College of Engg - Pre University Course, MPC – Mathematics, Physics, Chemistry, Akshara Junior College Public links: - github: https://github.com/dspraneeth07 - leetcode: https://leetcode.com/dspraneeth07 - codeforces: https://codeforces.com/profile/dspraneeth07 Profile highlights: - The best product I’ve worked on…: FlowOS — a full-stack agentic AI platform I architected from orchestration engine to autonomous execution, validated across 400 tests at 100%. - what i've built with AI: From a 3.1B-parameter LLM trained from scratch to sovereign OSINT, agentic platforms, local RAG systems, and AI products used at scale. - I'm the person in every team who…: Turns ambiguous problems into architecture, prototypes, production systems, and measurable outcomes then raises the technical bar for everyone. - What's hard for others but easy for me…: Going from a blank page to a working AI system - architecture, models, data, agents, infrastructure, and deployment without waiting for a playbook. Tal is a social hiring app where bosses and talent meet directly.