Veda Rama Charan Pisupati
Open to work1 years experienceAI/ML Engineer
AI/ML Engineer specializing in LLMs, RAG, and production-grade AI systems.
Standing Rank
Rank Not Available
Developer Badges
No badges earned yet
Skills & Technologies
29 skillsWork Experience
AI/ML Engineer Intern
Pratilipi
- Owned Setu and Vaani features end-to-end across LLM workflows, FastAPI/Next.js services, and production infrastructure, shipping multilingual AI capabilities used by internal product teams.
- Improved production multilingual adaptation quality by redesigning Setu into a 3-stage LLM pipeline, reducing defects by 58% across 12 languages, lowering cost to ~$0.27/chapter, and driving rollout through human and benchmark-based evaluation.
- Built and productionized a multilingual content-closure classifier using multilingual-e5, Logistic Regression, Metaflow, AWS Step Functions, and Grafana, processing ~6,000 reviews/day across 12 Indic languages with 90.5% cross-lingual accuracy.
- Reduced Setu pipeline latency by ~53% (43 min -> 20 min for a 10-chapter job) by introducing inter/intra-chapter parallelism, prefetching shared state, and fixing unsafe async/database access patterns in production.
- Re-architected Vaani authentication into a Backend-for-Frontend using Next.js + FastAPI, cutting Auth0 token fetches by 10-14x per page load, eliminating a 100-140 req/s self-traffic spike, and reducing CORS/CSRF surface.
- Built a server-to-server Comics Studio <-> Vaani integration using Auth0 M2M, scoped permissions, and automatic cost attribution, and wrote the integration documentation that enabled another product team to ship against the API.
Full Stack AI Intern
AI4Bharat (IIT Madras)
- Designed and implemented multi-tenant isolation for the platform, migrating 10 consumer-app tenants off a shared database onto per- tenant routing with isolated data, auth, and onboarding workflows on GCP.
- Shipped a multi-model speech-synthesis feature into production - built the request API, backend orchestration, and configuration UI (tone, voice, gender, etc.) that submits batch jobs to AI4Bharat's in-house and third-party audio models; adopted by users across tenants.
Projects
KhetAI - Agentic LLM Assistant for Indian Agriculture
- Built a tool-augmented LLM system for agriculture queries, with intent-classified routing to 6+ live API tools (Groq-hosted LLM, weather, government commodity prices via Agmarknet/data.gov.in, news, plant disease via PlantNet, geolocation); used dynamic model-size selection per query and a transparent multi-phase reasoning loop (Understanding -> Tools -> Analysis -> Response) surfaced to the UI.
- Validated with 10 farmers near my college, iterating on prompt design and tool-selection logic based on direct field feedback; supports Telugu, Hindi, and English; live as an Android APK today.
Telugu Text-to-Speech - Fine-Tuned SpeechT5
- Fine-tuned Microsoft's SpeechT5 on 8,576 Telugu audio samples from SPRINGLab's IndicTTS corpus, building a custom Telugu-to- Latin transliteration layer so the English-trained tokenizer could process Telugu script; integrated SpeechBrain x-vector speaker embeddings and HiFi-GAN vocoder for the full inference pipeline.
- Drove validation loss from 0.67 -> 0.45 over 1,000 training steps; open-sourced on Hugging Face (hf ), 613+ downloads to date.
FlowOSS - Local-First Voice Dictation App
- Built FlowOSS, a cross-platform, local-first voice dictation app for Linux and Windows in Rust/Tauri that performs private, fully on- device speech-to-text and inserts text into any active application via global shortcuts.
- Designed the system as a modular Rust workspace integrating Silero VAD, sherpa-onnx, and Parakeet for low-latency transcription, plus an assist workflow for highlighted-text Q&A and a desktop UI with tray, overlay, and settings support.
Leaderboard Standings
Leaderboard Position Pending
Global test scores, peer standing percentiles, and algorithm leaderboard ranks are updated dynamically.
Assessment Highlights
Assessments Not Completed
Coding evaluations, system assessment results, and conceptual score badges will appear here after taking a test.
AI Collaboration Score
AI Collaboration Score Pending
Developer coding behavior, assistant cooperation, and AI pair-programming indicators are evaluated during live coding sessions.
Role Compatibility Profile
Role Compatibility Analysis Pending
Custom matching reports, candidate role compatibility percentiles, and core engineer strength profiles are processed once conceptual code screenings are complete.
Achievements
NPCI Bharat AI Quest Finalist
Finalist in NPCI Bharat AI Quest, awarded AI Elite recognition.
ISRO Bharatiya Antariksha Hackathon Shortlist
Qualified for shortlist stage (top 140 teams) in a national-level space technology hackathon.
About Details
Professional Bio
AI/ML Engineer with experience in LLM workflows, FastAPI, and production infrastructure. Skilled in building multilingual AI pipelines, agentic systems, and speech synthesis features. Passionate about deploying scalable AI solutions for real-world impact.
B.Tech in Artificial Intelligence & Machine Learning
Vasireddy Venkatadri Institute of Technology (VVIT) (2022 - 2026)