3s → 1.2s
Real-Time Voice Infrastructure
Built LiveKit voice pipelines with streaming inference, connection pooling, and smart caching, reducing response time from 3 seconds to 1.2 seconds.
Voice and chat AI agents fail in ways nobody sees until a customer complains. I build VEE Legion, which finds those failures first with simulated callers and red-team attacks. I also build the real-time voice pipelines, retrieval and fine-tuned models behind them.
// Work · VoiceraCX
Independent QA for voice and chat AI agents
I defined the product and built the backend of VEE Legion. It tests AI agents the way customers do, with simulated callers and red-team attackers, then shows exactly where they break. Break the agent before your customers do.
01 · The problem
An AI agent can handle thousands of calls a day. A few go wrong, and nobody knows which until a customer complains. VEE Legion finds them first.
02 · The callers
Scenario-driven callers are frustrated, confused, anxious, pushy, tech-savvy or on a noisy line. Red-team callers try prompt injection, jailbreaks and social engineering. All of them test the agent exactly as it is, on VEE, Twilio, Retell, Vapi or Synthflow.
03 · The rules
Pass or fail can be a sentence, such as “verify the caller’s identity before changing account details”, judged by an LLM next to automated checks for WER/CER, latency, tool calls, PII and keywords. A failed judge run never counts against the agent.
04 · At scale
Slot-based orchestration in Go ramps load up instead of firing everything at once, so the test measures the agent and not the test rig.
200
concurrent calls verified
0
transport errors
05 · The why
Every call is scored and every failure is shown turn by turn. Real production calls can be imported and graded with the same checks, so tests and reality sit on one scale.
06 · Try it
Pick a caller and run the test. These conversations are illustrative, written for this demo, but the checks work the way Legion’s do.
Goal · Change my billing address, quickly.
Press “Run test” to start the call.
Illustrative conversations written for this demo, not real call data.
// Manifest_v5.0
Three layers of a voice agent that I have built and measured: how it talks, what it knows, and the models behind it.
The generic path of a voice agent. The solid stages are parts I built, across VoiceraCX, Alesa AI and the Orpheus TTS project.
3s → 1.2s
Built LiveKit voice pipelines with streaming inference, connection pooling, and smart caching, reducing response time from 3 seconds to 1.2 seconds.
−90%
Migrated Chroma to Qdrant with hybrid dense-sparse search, reducing query latency by 90% for voice-agent workloads.
10k+ terms
Fine-tuned Llama-3.1-8B with PEFT on a 10,000+ term dream-interpretation dataset and trained Orpheus TTS to generate SNAC audio tokens.
// Work · Education
Drawn like a commit graph: my work on the main line, and my degree running alongside it as a branch that merged in July 2026.
June 2026 — Present · Pune
AI R&D Engineer · VEE Legion
Building VEE Legion, an independent QA platform that tests voice and chat AI agents with simulated callers, automated scoring and red-team attacks.
June 2025 — June 2026 · Pune
Real-time voice pipelines and scalable vector retrieval for voice-agent products.
Feb 2025 — June 2025 · Remote
Svelte web application features in a production workflow.
Nov 2024 — Mar 2025 · Remote
Tangent Mind, a tarot, horoscope and dream-interpretation platform.
Nov 2022 — Jul 2026 · Pune
Bachelor of Engineering in Computer Engineering · Honors in Data Science
Manifest_v7.0
Seven projects, from voice models to company memory. Open any tile for its architecture and design decisions, or ask Wisp to explain it.
Manifest_v7.0
Seven projects, from voice models to company memory. Open any tile for its architecture and design decisions, or ask Wisp to explain it.
Explore the projectsFalkorDB
October 2026
A company brain that knows who's allowed to know what
7
Connectors
13
Graph node types
12
Agent tools
Design questions it answers
Generated by the model
Llama 3.2 3B
March 2026
A language model where the 'language' is sound
19min
1 epoch, 299 steps, single GPU
FastAPI
July 2025 – June 2026
Offline-first AI school ecosystem for Tier-2 and Tier-3 India
4
Admin, Teacher, Parent, Student


Next.js 16
May 2026
A gentle daily companion for mood, water, meals and tiny wins
60sec
Mood, water, meals, sleep and tiny quests in about a minute
LLM
April 2025
Intelligent agricultural planning from seed to sale
1st
Winner at Devclash 2025, DY Patil




Flask
2025
AI-based defense news aggregator with sentiment analysis and summaries
10+
Global defense publishers aggregated in real time


BERTSum
March 2025
AI-powered legal document simplification and Q&A
1st
Winner at Synapse 2.0, CCOEW
// System_Status: Operational
The tools I build with, grouped by what they are for.The tools I build with, and how they connect. Hover a logo or a node to see what connects to it, or drag the graph.
// Recognition
1st prize
24-hour hackathon at DY Patil Pimpri
1st prize
AI hackathon at MKSSS CCOEW
1st prize
24-hour hackathon at COEP Mindspark 24
99.35 percentile
Merit rank 971 among 400,000+ candidates
// Contact
Tell me about the role, the project or the idea. The quickest way is email, and the form below opens your mail app with everything filled in.
Pune · UTC+05:30
Currently building VEE Legion at VoiceraCX.