About me
AAII EEnnggiinneeeerr wwiitthh 22..55++ yyeeaarrss ooff eexxppeerriieennccee aanndd aa ssttrroonngg ffuullll--ssttaacckk bbaacckkggrroouunndd iinn RReeaacctt,, NNeexxtt..jjss,, TTyyppeeSSccrriipptt aanndd NNooddee..jjss.. II bbuuiilldd pprroodduuccttiioonn AAII ffeeaattuurreess:: aa RRAAGG aassssiissttaanntt wwiitthh ppggvveeccttoorr sseemmaannttiicc sseeaarrcchh,, aa ttooooll--ccaalllliinngg AAII aaggeenntt wwiitthh hhuummaann aapppprroovvaall,, aann LLLLMM eevvaalluuaattiioonn ssuuiittee,, AAII ccaallll ttrraannssccrriippttiioonn aanndd aannaallyyssiiss,, aanndd aa rreeaall--ttiimmee ddeeeeppffaakkee ddeetteeccttiioonn ssyysstteemm ffoorr vviiddeeoo iinntteerrvviieewwss.. II ffooccuuss oonn ssaaffee aanndd tteessttaabbllee AAII —— vvaalliiddaatteedd LLLLMM oouuttppuuttss,, PPIIII mmaasskkiinngg,, aauuddiitt llooggss,, aanndd ffaallllbbaacckkss wwhheenn aa mmooddeell ffaaiillss.. MMyy ffuullll--ssttaacckk sskkiillllss lleett mmee oowwnn aann AAII ffeeaattuurree eenndd ttoo eenndd,, ffrroomm tthhee mmooddeell ccaallll ttoo tthhee UUII aanndd ddeeppllooyymmeenntt..
Experience
Working on full-stack projects and building AI chatbots, RAG and Generative AI features — on a Next.js, TypeScript, Node.js and PostgreSQL stack.
- Full-stack projects: built SalesRocks features (multi-tenant setup, two-factor login, in-browser calling, dashboards), Atna Product modules and the company website.
- AI chatbot: built Ask AI — a chat assistant that streams answers, cites sources and suggests follow-up questions.
- RAG: built the vector search pipeline — data is chunked, embedded with Gemini embeddings and searched with PostgreSQL + pgvector.
- Generative AI: built Do AI, an AI agent with LLM function calling, and AI call analysis with Gemini + Claude structured outputs.
- Safe AI: LLM eval suite (32 + 10 checks), PII masking, Zod validation, audit logs and model fallbacks.
- AI video security: built the Deepfake Meeting Guardian for Google Meet, Teams and Zoom interviews.
Built high-performance dashboards for business-critical workflows.
- Built scalable web apps with React.js, Next.js and TypeScript for business-critical dashboards.
- Created reusable component libraries that sped up development across multiple projects.
- Integrated REST APIs for real-time data visualization and decision-making.
- Turned Figma designs into responsive, accessible UIs with Tailwind CSS.
Built responsive interfaces and integrated REST APIs for production features.
- Developed responsive user interfaces with React.js, JavaScript, TypeScript, HTML5 and CSS.
- Improved performance with component optimization and efficient state management (Hooks, Context API).
- Worked with designers and backend teams to ship production-ready features.
Skills
Projects
- Ask AI (RAG): answers questions from the company's own calls, emails, leads and deals — streams answers, cites sources and suggests follow-up questions
- Vector search: records are chunked, embedded with Gemini embeddings and stored in PostgreSQL + pgvector, indexed on save plus a 5-minute background job
- Do AI (agent): uses LLM function calling to read data and propose actions (create task, draft email, score lead) — nothing runs until a person approves
- Safe AI: every AI action is checked with Zod twice and logged for audit; PII is masked before it reaches the LLM
- LLM evals: 32 known-answer checks + 10 system checks, run with one command, to catch mistakes when prompts change
- AI call analysis: Gemini transcribes calls; Claude returns structured JSON — summary, sentiment, buying intent, objections, next steps
- Switchable Gemini / Claude provider with safe fallbacks when a model or quota fails
- Pluggable detection: frames and audio go to detector model servers, with a local TensorFlow.js model and audio heuristics as fallback
- Iris-based eye-gaze tracking with MediaPipe to detect look-aways and tab switches — nothing to install for the candidate
- 3 capture methods: a server-side meeting bot (ffmpeg), browser screen capture, and a Chrome extension
- Live audit dashboard with real-time risk scores, per-participant risk reports and bulk 'scan all deepfakes'
- Built the dashboard, customer management, profile and lookup modules
- Built shared platform-admin UI components used across the product
- Added real-time notifications with Server-Sent Events (SSE)
- Built with Next.js and TypeScript for a fast, clean user experience
- AI job matching: matches a resume to jobs and gives a match score
- AI resume parsing: reads resume files and pulls out skills and experience
- AI auto-answers: fills common application questions automatically
- AI candidate screening: checks and ranks candidates for employers
- Role-based access for Job Seekers, Employers and Admins with NextAuth; resume uploads via Cloudinary
- Designed and built a responsive website aligned with GRB brand identity
- Built fast, clean UIs with React.js and Tailwind CSS for desktop and mobile
- Integrated SendGrid contact and enquiry forms to automate lead management
Education
Kumararani Meena Muthiah College of Arts & Science · Chennai, India
2018 — 2021Santhome Higher Secondary · Chennai, India
Besant Technologies · India
Contact
Looking for a React / Next.js developer who understands APIs, Redux state management, and production deployments? Let's connect.