July 2026 – Present · Bangalore, India (On-Site)
Project: Security Engineering — AI for Security & Security for AI
Converted from the cybersecurity internship into a full-time Security Engineer at the IS2 grade on 16 July 2026. Working across AI for Security and Security for AI — autonomous detection tooling and ML/agent supply-chain hardening — alongside VAPT engagements, SOC detection engineering across Microsoft Sentinel (KQL) and Splunk, and a shift-left DevSecOps focus.
Delivering reusable, explainable detection engineering and maintaining an open-source security tooling suite; supporting GRC control audits (J-SOX, MICS) across IT general controls — access, change, and segregation-of-duties; and presenting security architecture and business value to executive leadership.
January 2026 – July 2026 · Bangalore, India (On-Site)
Project: Agentic Email Security
Completed intensive cybersecurity training — OSI/TCP-IP, DMZ design, VAPT (Burp Suite, Nmap, sqlmap, Nessus), OWASP Top 10 remediation (SQLi, XSS, Broken Access Control), SOC operations (Microsoft Sentinel, Splunk SPL, Defender EDR/XDR), AWS Shared Responsibility Model, MFA/SSO (SAML/OAuth 2.0), DevSecOps shift-left, and OS hardening. Engineered a multi-agentic AI email-security system of 7 LangGraph-orchestrated agents to autonomously detect and neutralize threats in real time.
The system reached ~96–97% threat-detection accuracy with a <0.8% miss rate and ~1.3% false positives, and cut SOC email triage from ~30 minutes to a few seconds via an explainable, SOC-ready decision pipeline — leading to conversion into a full-time Security Engineer (IS2) role at the end of the internship.
May 2025 – July 2025 · Shantou, China (Remote)
Project: Gated Multi-Domain CUT Model
Designed a novel Gated Multi-Domain CUT model with attention-based architecture and hybrid contrastive loss for high anatomical fidelity.
Outperformed baselines (CycleGAN, StarGAN, CUT) achieving SSIM of 0.9858.
May 2025 – July 2025 · IIT Ropar, Punjab (Remote)
Project: Sanchalak
Built FastAPI backend with Prolog-based rule engine, end-to-end pipeline using Whisper, Azure Translator/TTS, and Gemma LLM for personalized voice interactions.
Achieved <5s latency per query with 100% deterministic eligibility checks via hybrid Prolog+LLM architecture.
July 2024 – February 2025 · Chennai, India (On-Site)
Project: Medical Audio Classification
Fine-tuned CLAP and CLIP on 20k+ medical audio samples from PhysioNet, BHIC, AudioSet, VGG-Sound. Built annotation + preprocessing pipeline with t-SNE visualization.
Enabled zero-shot analysis across 10+ body sound classes with, high embedding quality.
June 2024 – August 2024 · Chennai, India (Hybrid)
Project: HAR-GCNN
Developed HAR-GCNN model using Graph Convolutional Networks on PAMAP2 sensor dataset with robustness to 66% missing labels.
Achieved 99.99% accuracy, outperforming CNN (99.75%) and LSTM ,(98.10%) across 3-25 activity classes.
May 2023 – June 2024 · Chennai, India
Project: ML Research & Development
Participated in research initiatives, collaborated on ML projects, and contributed to team knowledge base.
Built foundation in applied ML research and collaborative development.