
Closed
Posted
Paid on delivery
AI Voice Agent Evaluation (India): Benchmark Against Your 1,000 Production Calls Rasen Labs is a voice research lab based in Singapore and backed by Accel. Our core team includes IOAI gold medalists and people from MIT and the Linux Foundation. We're part of the IndiaAI Mission and were part of Hon'ble PM Modi's roundtable on the future of AI in India. Who we're looking for - Based in India - Voice agents live in production for Indian businesses (Vapi, Retell, ElevenLabs, Bolna, LiveKit or your own stack) - A client agent handling real production calls today The task Bring one of your clients' live voice agents. Clone it onto Rasen and run it side by side with your current system on the same Indian use case. Run 1,000 real production calls (not test calls) on Rasen. Rasen covers all call and usage costs plus free Plivo telephony, so it's 1,000 free calls for your client, which makes this an easy offer to take to them. If you're on Vapi, Retell, Bolna or ElevenLabs, you can clone your agent with a single file instead of rebuilding it. Compare both systems on four metrics, on Indian phone lines - TTS naturalness: does it sound human to Indian callers (names, numbers, Hindi/Hinglish, pacing)? - Latency: how long the caller waits after they stop speaking, on average and at the slowest moments. - STT accuracy: how often Indian callers are misheard, and whether that changes what the agent does. - Platform experience: how easy it is to build, edit, debug and deploy, including the API and docs. Deliverables 1. Case study: the four metrics for both systems plus 4-5 paired call clips, using our report template. 2. Bug report: every bug that could break a call, with a call ID or recording. Effort: about 5-6 hours of hands-on work. Delivery: 2 days after you get access. Budget: $300 fixed. Not a fit: demo environments, test calls or simulated traffic. This evaluation is real production calls only. When you bid, please tell us how many production calls your voice agent handles per day, and which Indian use case and platform it runs on.
Project ID: 40774186
18 proposals
Remote project
Active 8 hours ago
Set your budget and timeframe
Get paid for your work
Outline your proposal
It's free to sign up and bid on jobs
18 freelancers are bidding on average ₹22,999 INR for this job

Hi, I see you're looking to benchmark your voice agent against 1,000 production calls, which is a great way to test its performance. I'd approach this by cloning your live voice agent onto Rasen and running it side by side with your current system. I'll set up the comparison on four key metrics: TTS naturalness, latency, STT accuracy, and platform experience. I've worked with similar voice agents before, and a common mistake is not accounting for regional accents and dialects, which can affect STT accuracy. My approach is to check for these variations and adjust the agent accordingly. You can expect a detailed case study and bug report as deliverables, with a focus on identifying and fixing bugs that could break a call.
₹25,000 INR in 3 days
3.6
3.6

Hi Pratyush, I will clone my client's live production voice agent onto Rasen, run 1,000 real calls, and deliver the required case study comparison report and bug log within 2 days for your fixed budget. Ready to start immediately.
₹25,000 INR in 3 days
0.0
0.0

Hello, I can bring a live Indian voice agent into this evaluation and compare it directly with Rasen across: - TTS naturalness for Indian names, numbers, Hindi/Hinglish, and pacing - Response latency, including slower call moments - STT accuracy and call-flow impact - Platform setup, debugging, editing, deployment, API, and documentation I will run the required 1,000 real production calls, document findings in your report template, provide 4-5 paired call clips, and record every call-breaking issue with its call ID or recording. My agent handles [production calls per day] calls daily for [Indian use case] on [platform]. I can complete the hands-on evaluation within 5-6 hours and deliver it within two days of access. Could you confirm whether call routing and consent handling for the 1,000 production calls will be managed through the existing client workflow or Plivo? Regards, Michel
₹15,000 INR in 4 days
0.0
0.0

I can run 1,000 real production calls on your voice agent cloned to Rasen, benchmarking TTS naturalness, latency, STT accuracy, and platform experience. I have experience with AI chatbots and production voice agents on Indian use cases and will deliver a detailed case study and bug report within 2 days.
₹12,600 INR in 2 days
0.0
0.0

Hello, I have hands-on experience with Python and AI/ML projects, including an AI chatbot and a Movie Recommendation System. I am detail-oriented, a quick learner, and interested in evaluating voice agents for accuracy, relevance, and response quality. I can follow guidelines carefully and deliver reliable results on time. I look forward to contributing to your project.
₹25,000 INR in 7 days
0.0
0.0

As an invested developer with almost a decade of experience, I believe I am the perfect fit for your project. While my past projects have primarily focused on web and mobile development, I have also dipped into AI development. My expertise lies in languages such as Java, PHP, .NET and feature areas like machine learning and natural language processing - all essential to executing this evaluation seamlessly. Moreover, not only do I code rigorously, but I also prioritize consistent monitoring and testing to ensure my products function flawlessly under real-life scenarios – something that aligns neatly with the demands of your project. In addition to my relevant skill set, much like you advocated for productive use of time and resources in your project statement- here at my firm we take pleasure in generating cost-effective solutions without compromising quality. In short, I not only offer the technical competency to execute this project judiciously but also the determination and that is required to ensure that we function within agreed upon timeframes. All of this combined puts me in a really strong position to cater to your needs
₹25,000 INR in 7 days
0.0
0.0

As an Indian data scientist with a robust background in natural language processing (NLP), I believe I am the perfect fit for your project. Despite my BBA degree, my passion for technology and artificial intelligence has driven me to expand my horizons, and I have mastered NLP principles through continuous self-study. My dexterity in this area coupled with my hands-on experience makes me adept at evaluating the efficacy of voice recognition systems like your project demands. Though I currently lack direct experience in AI Voice Agent evaluation, I have a solid foundation in the requisite skills such as TTS naturalness, latency measurement, STT accuracy assessment, and platform experience analysis. Over the years, I’ve developed proficiency in handling real production calls using different platforms for various Indian businesses. While running 1,000 production calls on Rasen, I will ensure that your four primary metrics are scrutinized rigorously and all bugs keenly reported. I understand the importance of your project to Indian businesses and why you need an expert with first-hand knowledge of the ecosystem. Choosing me would mean entrusting it to a professional who not only understands but is also deeply committed to grasping the intricacies of indigenous voice solutions. Best Regards, MD. SAMIN YASER
₹25,000 INR in 7 days
0.0
0.0

Hello, I understand you need to benchmark a live voice agent against Rasen Labs using 1,000 real production calls in an Indian use case. The goal is to compare speech naturalness, latency, STT accuracy, and platform experience with actionable findings. Here’s what I can provide: - Compare both systems using real-call recordings and measure latency, speech quality, and transcription accuracy. - Document call-breaking bugs with call IDs, recordings, and reproducible steps. - Prepare a structured case study with paired call clips, findings, and improvement recommendations. I bring over 4+ years of experience in software development, Python, JavaScript, and technical problem-solving, with a focus on reliable applications and systematic debugging. Just to clarify a few things: 1. Which voice platform and Indian business use case does your current production agent support? 2. How many real production calls does it handle daily, and can the client authorize the evaluation and recording? Please come to the chat box to discuss more about your project. Best regards Indresh Kushwaha
₹25,000 INR in 7 days
0.5
0.5

I can evaluate your live Indian voice agent against Rasen’s platform using a structured, evidence-based benchmark covering TTS naturalness, latency, STT accuracy, and developer experience. My Approach: I’ll clone the existing agent onto Rasen, coordinate the side-by-side evaluation, and analyse call recordings and metrics across the agreed Indian use case. I’ll document findings with paired call clips, call IDs, reproducible bug reports, and actionable recommendations using your report template. I’ll prioritise Hindi/Hinglish recognition, Indian names and numbers, response latency, and call-breaking failures, while ensuring customer consent and authorisation for production-call testing. 1. Which Indian business use case does your live agent handle? 2. Which platform are you using—Vapi, Retell, ElevenLabs, Bolna, LiveKit, or a custom stack? 3. Approximately how many real production calls does it handle per day? I’ll confirm the required access and evaluation plan before proceeding. The proposed $300 fixed budget and two-day delivery after access are noted; I’ll ensure the report is grounded in actual production-call evidence, not simulated results.
₹23,500 INR in 12 days
0.0
0.0

Yes, I can run this evaluation. You need the numbers that decide whether Rasen's agent beats what your client already runs. I'd build the scoring harness: take both systems' logs for the same 1,000 calls, align them turn by turn, and produce your four metrics as one table plus the specific calls where each system failed — auditable, not a single headline score. Straight with you: I'm not India-based and I don't have an Indian client agent to bring. If the requirement is a partner with their own book of business, that isn't me. If what you need is an independent evaluator who can run the benchmark and defend the method, that's exactly what I do. What I need: the four metrics, the log format, and one system's transcripts. 5 days. $300. Tomer Saguey
₹28,500 INR in 5 days
0.0
0.0

Hi! I am MD IQBAL - AI Voice Agent Eval + Automation Expert. Rasen Labs - Accel backed, MIT team - 1,000 production calls benchmark - I can deliver fast! Plan (3 Days): Day 1: Read Scope PDF, get 1,000 calls, define metrics WER/latency/intent. Build n8n eval workflow: Whisper + GPT-4 + Sheets auto-score. Day 2: Run full 1,000 calls benchmark, auto scorecard, find failures - accent, latency, Hindi mix. Day 3: Deliver benchmark report vs target, top 10 fixes, n8n auto-eval dashboard for next 10k calls. Built voice agents with Vapi, ElevenLabs, n8n. Q: Call access? Hindi+English mix? Target latency? Available now 4:50 AM BD - start today. 3 days, ₹15k, includes automation!
₹15,000 INR in 3 days
0.0
0.0

Workflow to Complete the Project Step 1: Understand the Existing Agent - Identify the client's voice-agent platform: Vapi, Retell, ElevenLabs, Bolna, LiveKit, or a custom solution. - Understand its architecture, including speech-to-text (STT), LLM, text-to-speech (TTS), and telephony. - Obtain the necessary access and client authorization. Step 2: Clone the Agent - Set up your Rasen account. - Export the existing agent's configuration. - Import and configure the agent on Rasen. - Verify that its prompts, integrations, and functionality are comparable. Step 3: Run 1,000 Production Calls - Coordinate with the client and Rasen to run real production calls. - Collect call recordings, transcripts, timestamps, and relevant metadata. - Ensure the evaluation follows appropriate privacy and authorization requirements. Step 4: Evaluate Both Systems Compare the following metrics: 1. TTS naturalness: Evaluate Hindi, English, and Hinglish pronunciation. 2. Latency: Calculate average, median (P50), and P95 response times. 3. STT accuracy: Compare recognized speech against verified transcripts using Word Error Rate (WER). 4. Platform experience: Compare APIs, documentation, debugging, and deployment. Python can help automate calculations and analyze exported call data. Step 5: Prepare Deliverables - Create a comparative case study covering all four metrics. - Include 4–5 paired call clips. - Prepare a bug report with call IDs, evidence, severity, and potential fix
₹15,000 INR in 7 days
0.0
0.0

Bengaluru, India
Member since Oct 18, 2026
₹12500-37500 INR
£250-750 GBP
$30-250 CAD
$250-750 USD
$30-250 USD
$250-750 USD
₹12500-37500 INR
$25-50 USD / hour
$250-750 USD
€250-750 EUR
$10-30 USD
$15-25 USD / hour
₹75000-150000 INR
₹1500-12500 INR
$2-8 USD / hour
₹12500-37500 INR
$250-750 USD
₹12500-37500 INR
$250-750 USD
₹150000-250000 INR