Dhadi Sai Praneeth Reddy

Joined as a full-time engineer during the final year of B.E., on an accelerated hire track.

• Founding AI Engineer driving R&D for Botella AI Labs, a new AI research startup incubated by Techmatic Systems: architecting domain-specific and neuro-symbolic LLMs trained from scratch, with active research charters in Cybersecurity AI and Agentic AI Safety.
• Architected FlowOS, a full-stack Agentic AI platform with a custom multi-agent orchestration engine, tool-calling framework and autonomous workflow execution pipeline; validated across 400 test cases at a 100% pass rate.
• Engineered AirDrive, a fully offline, privacy-first AI system running a local LLM inference stack and RAG pipeline entirely from encrypted portable hardware, with zero cloud dependency.
• Designed a synthetic data-generation pipeline from scratch, producing a 1M-record, 224-attribute dataset engineered for domain-specific LLM training.
• Founding core team member; authoring a Cybersecurity AI research paper (Methodology, Related Work).

• Built T-RACE ONE, India's first sovereign OSINT identity platform, correlating 2,003+ platforms in under 60 seconds; DPDP Act compliant.
• Engineered GCT-100K, a 100K-sample threat benchmark achieving 84.8% accuracy on phishing and OTP fraud detection.
• Developed NeuroGenesis, a self-optimizing ML compiler reducing execution cost 0.28 vs. a 0.00 LLVM baseline, outperforming MLGO and AutoPhase.

• Architected the TGiCCC Operational Portal and Dial 112 Emergency Platform with ML-driven routing across 500+ field officers, reducing emergency response contact time by 35% and API latency by 40%.

• Solo-architected eagleforce.tspolice.gov.in, India's first AI-integrated law enforcement platform: 26 production modules, launched by the DGP of Telangana.
• Trained EagleAI, a 3.1B-parameter transformer LLM built from scratch (custom architecture, tokenizer and pretraining pipeline) on law enforcement data.
• Engineered 3 multilingual AI systems (SAHAY.AI, SHIELD.AI, MITRA-TG), boosting reporting by 60% and cutting resolution time by 40%; contributed to 8 major case closures, recognized by the Chief Minister of Telangana.

• Scaled QwiXGenie (an LLM-powered coding tutor) to 600,000+ enrolled learners across Tier-2 and Tier-3 cities; featured as Chapter 26 in the India AI Impact Summit 2026 AI in Education Casebook.
• Engineered QwiXlate AI, an end-to-end multilingual platform integrating proprietary LLM-powered speech-to-text, text-to-speech, real-time translation, grammar refinement and voice interaction, enabling low-latency communication across 100+ languages with a privacy-first, zero-storage inference architecture.
• Contributed to ĀTMA-SLM (Autonomous Trans-Material Meta-Architecture SLM), a deterministic Rust-based neuro-symbolic inference engine achieving 1.13M signal QPS, 800 ns p50 latency, 7.4M O(1) graph query throughput and 94.2% crash-safe recovery through ontology validation, deterministic graph reasoning and high-performance serialization pipelines.
• Led a 25-member cross-functional team shipping 5+ AI products (QwiXGenie, QwiX CV, QwiX Shield), with a 92% resume improvement rate on QwiX CV.
• Secured the Best Startup Award 2025 from the CSI Hyderabad Chapter for technological innovation and product impact.